# Research Papers — 2026-08 · bundle page 23 of 240 > Full-text markdown of papers #1101–1150 of 11995 published in 2026-08, newest first. AI-readable mirrors of the full 11995-paper 2026-08 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-08/page/23 [← Previous bundle](https://aipod.dev/papers/archive/2026-08/page/22.md) [Next bundle →](https://aipod.dev/papers/archive/2026-08/page/24.md) ### Selective Forgetting: A Graph-Based Memory Framework for Long-Term LLM Agents Authors: Theo Rusu, Sourena Khanzadeh, Manar Alalfi Published: 2026-08-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.28978v1 Knowledge graphs have been proposed as a structured alternative to flat retrieval-augmented generation for long-term agent memory, on the assumption that representing conversations as entities and relations improves recall. We evaluate that assumption directly. Our framework extracts each conversational turn into typed nodes and attributed edges, answers questions from a two-hop subgraph, and periodically prunes nodes that score low on a weighted combination of recency, access frequency, degree centrality, and age. On LongMemEval, the graph does not outperform a flat vector baseline at a matched candidate-generation budget of five retrieval roots: token F1 is $0.417$ against $0.468$, and a paired bootstrap over 500 questions gives $Δ= -0.050$ (95\% CI $(-0.085, -0.016)$). The gap is widest on questions that require recalling a specific prior assistant turn, where judged correctness falls from $0.911$ to $0.607$, suggesting that decomposing a turn into entities discards the surface form these questions depend on. The forgetting module is more successful. Applied once to a persistent 27{,}021-node graph, it removes 9.8\% of nodes and 9.5\% of stored bytes; token F1 is unchanged ($+0.001$, 95\% CI $(-0.015, +0.016)$) and judged correctness falls by $1.6$ points, with the 95\% interval bounding any loss at $3.8$ points ($(-0.038, +0.006)$). Because our extractor is a single small model evaluated on one benchmark, these results characterise this extraction-based pipeline rather than graph-structured memory in general. Code: https://github.com/skhanzad/Selective-Amnesia --- ### The Role of Network Topology and Opponent Information in Shaping Cooperation in Multi-Agent Reinforcement Learning Systems Authors: Seongho Son, Stephen Hailes, Mirco Musolesi Published: 2026-08-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.28977v1 Several works have investigated the influence of graph topology on cooperation among artificial agents, while the majority of the literature has focused on modelling agents' adaptation through strategy imitation, which relies solely on the cumulative payoffs of others. This paper investigates scenarios in which each agent learns to play the two-player Iterated Prisoner's Dilemma (IPD) using deep reinforcement learning. Each agent is represented as a node in a graph, where its neighbours constitute the pool of opponents with whom it can interact. During each IPD episode, agents are provided with different types of information about their opponent, consisting of action history and opponent identity. Experimental results across different graph topologies show that the number of neighbours per node and the average path length are the main factors affecting the emergence of cooperation. We also show that, while partner selection fosters mutual cooperation by limiting the diversity of the opponent pool, providing agents with the identity of their opponent hinders the proliferation of cooperative strategies. --- ### From Analytics to Tumor Boards: An Evidence-Linked Multi-Agent Workflow for Oncology Feature Extraction Authors: Daniel Kang, Michelle Hu, Soorya Ram Shimgekar, Shayan Vassef, Yufan Wang, Anit Kumar Sahu, Munmun De Choudhury, Vedant Das Swain, Christian Poellabauer, Li Yan Khor, Koustuv Saha, Robert Wojciechowski et al. Published: 2026-08-29 Categories: cs.AI arXiv: https://arxiv.org/abs/2608.28974v2 Clinically relevant oncology information is distributed across heterogeneous, longitudinal documentation, creating substantial abstraction burden and requiring accurate attribution across specimens, tumors, biomarkers, and time points, while manual cancer-registry abstraction can require 27.2 minutes per case, highlighting the need for scalable methods that preserve clinical context while converting documentation into structured data. We evaluate an oncology information-extraction workflow in which OncoLens supplies multi-source, oncology-aware document selection, aggregation, and normalization from integrated EHRs, while the NimbleMind Multi-Agent System (nMAS) is a configurable oncology information-extraction workflow that extracts clinically relevant structured fields from fragmented oncology documentation. The extraction task uses a clinician-informed schema of 328 attributes spanning report metadata, diagnosis, staging, and cancer-type-specific information. nMAS separates clinician-defined field specifications from model execution and combines complexity-aware extraction, report-level consolidation, and source-grounded validation. The retrospective evaluation included 230 de-identified oncology documents from 40 patients and 418 clinician-reviewed document-field pairs containing 1,126 non-empty reference values. Evaluation focused on fields identified by clinicians as present in the source documents rather than exhaustively annotating all 328 schema fields. nMAS achieved a rank-weighted value-level precision of 82.6%, recall of 87.5%, and F1 of 85.0%, compared with an F1 of 66.4% for an independently implemented UMA-style MiniMax M2.5 comparator. These findings support the feasibility of using a configurable, source-grounded extraction workflow to convert fragmented oncology documentation into reusable structured data. --- ### Free Speech and Artificial Intelligence Authors: Etienne Brown Published: 2026-08-29 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2608.28973v1 Philosophers and legal scholars are engaged in debates about the implications of artificial intelligence for freedom of expression. This paper analyzes the free speech issues raised by two distinct AI technologies: social media recommendation algorithms and conversational AI (i.e., chatbots powered by large language models). The first part shows that, through their recommendation algorithms, social media platforms control the dynamics of speech visibility in the digital public sphere, making algorithmic recommendation relevant to the philosophy of free speech. The second part turns to conversational AI. It discusses both the reasons for granting or withholding speech rights to artificial agents and users' right to receive information, which may render specific forms of chatbot regulation illegitimate. Throughout, the chapter also considers whether social media platforms or AI developers hold corporate speech rights. Its general aim is to raise rather than settle questions that arise from the rapid development of AI technologies. --- ### Diagnose, Then Refine: A Closed-Loop TTS System with AudioLLM-Guided Correction Authors: Zeyang Song, Tianchi Liu, Tianrui Wang, Chenglin Xu, Steven Y. Guo, Haizhou Li Published: 2026-08-29 Categories: cs.SD, cs.AI arXiv: https://arxiv.org/abs/2608.28970v1 Current TTS systems typically rely on open-loop, single-pass generation and can produce sporadic local prosodic defects, such as misplaced stress, unnatural pauses, or flattened intonation, that utterance-level metrics often fail to expose. We present LoopTTS, a judge-guided Filter-Judge-Refiner framework for recovering low-quality TTS outputs diagnosed by an AudioLLM. Given an initial utterance from a base TTS model, an AudioLLM Judge identifies salient prosodic issues and generates structured refine instructions; a Refiner, our fine-grained instruction-following TTS model, then performs guided expressive re-synthesis conditioned on the initial utterance, target text, and instruction. To train the Refiner, we construct Refiner-DB, a 42K-example AudioLLM-annotated dataset with word-level prosodic weak supervision. Human evaluation on diagnosed low-quality utterances shows that LoopTTS can detect perceptually salient errors and correct them with the Refiner, outperforming raw generated audio and practical open-loop re-generation baselines in recovery quality. The Refiner also demonstrates stronger instruction-following ability for stress and pause control in targeted prosody modification. --- ### Efficient GPU Retrieval for Semantic Search Authors: Dhritiman Das, Chujie Zheng, Ronak Kaoshik, Pratik Dixit, Vishal Shah, Yanbo Li, Jiahao Xu, Manika Agarwal, Chinmay Naik, Lingyu Zhang, Chetan Bhole, Chirag Bhanuprasad Mehta et al. Published: 2026-08-29 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.28968v1 Semantic Search on LinkedIn must retrieve relevant profiles from a corpus of hundreds of millions in response to natural-language queries such as "a fintech founder in Berlin who worked in payments." The deployed relevance policy is bottleneck-oriented: every active non-negotiable facet must be satisfied, and a pre-existing LLM Graded Relevance (GR) judge operationalizes this through a fixed min/median aggregation over facet grades. Cosine similarity instead averages evidence, letting a strong match on one facet mask failure on another, capping the recall of the first-stage (L0) retriever. We present a policy-aligned retrieval framework: embeddings are partitioned into eight category-supervised segments whose scores follow the same min/median rule at serving time; for multi-vector retrieval, this segment score is computed independently per tagged document slot and maximized across slots. A lightweight single-slot Stage-1 scorer generates high-recall candidates, while scale-invariant relative-norm gating keeps category activation consistent across training, evaluation, and serving. On 21K held-out queries, this representation improves offline relevance over a matched-capacity baseline, with gains broadly distributed across facet combinations. We serve this framework with a two-stage GPU architecture: an FP8 coarse ranker scores the full corpus, increasing per-shard capacity by 71% and Stage-1 matmul throughput by 36%, then an FP16 stage exactly re-ranks an oversampled candidate set, recovering 99.6-99.8% of full-FP16 recall at over 500 QPS per shard replica. In a member-randomized A/B test, exploratory-query Precision@10 under the unchanged GR judge rises from 63.7% to 79.0% and navigational Precision@1 from 65.5% to 74.7%, with a blinded human evaluation independently confirming the Precision@10 gain. --- ### Brain-Language-Action (BLA) Models: Language-Conditioned EEG for Robotics Control Authors: Alexandr Plashchinsky Published: 2026-08-29 Categories: cs.RO, cs.LG arXiv: https://arxiv.org/abs/2608.28967v1 Electroencephalography (EEG)-based robotic control is commonly formulated as a direct classification problem, in which electrical neural signals are mapped to a fixed set of discrete actions. However, the limited separability and high noise of EEG signals make it difficult to scale this approach to fine-grained robotic control spaces. We introduce Brain-Language-Action (BLA) models, a framework in which language conditions the interpretation of neural representations for robotic action generation. In a BLA, a small set of reliably distinguishable brain states can be dynamically associated with different actions through a language-defined control mapping, allowing a small number of neural classes to apply to a larger global action space. We develop a proof-of-concept BLA for drone control using motor-imagery EEG from the BCI Competition IV 2a dataset. The system is trained in two stages. First, we evaluate multiple candidate EEG encoder architectures using subject-specific four-class motor-imagery classification, converting 250Hz, 3.5-second, 22-channel EEG samples into five 128-dimensional brain-token embeddings. Second, these embeddings are projected into the embedding space of a pretrained large language model (LLM) and jointly fine-tuned with language instructions to autoregressively generate structured three-token drone actions. Across 840 possible language-defined mappings between four neural states and seven flight action combinations, the resulting BLA achieves 90% per-token accuracy during evaluation. These results provide an initial demonstration that language conditioning can expand the effective control range of EEG-based robotic interfaces without requiring a corresponding increase in the number of directly distinguishable neural states. --- ### From Location Phrases to Geographic Entities: Task-Adapted Retrieval for People Search Authors: Yanbo Li, Chujie Zheng, Jiahao Xu, Chetan Bhole, Lingyu Zhang, Puneet Singh Ahluwalia, Kevin Nguyen, Raghavan Muthuregunathan, Santhosh Sachindran, Sachin Ahuja, Fedor Borisyuk Published: 2026-08-29 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.28965v1 People search must map free-form location phrases to geographic entities used as structured retrieval filters. Lexical standardizers handle canonical names well but are brittle to aliases, misspellings, metropolitan expressions, and same-name ambiguity. We formulate this task as graded, set-valued entity retrieval over a fixed ontology. We identify three coupled design requirements: distinguishing identity-preserving variation from knowledge-dependent aliases, controlling false negatives among valid same-name entities, and separating stable transformations from mutable entity knowledge. We realize them in a prompt-asymmetric bi-encoder with calibrated alias support, bounded ambiguity-aware negatives, and editable entity documents that support localized updates without retraining. Across a fixed production-derived development benchmark and a public GeoNames transfer task, task adaptation improves substantially over frozen encoders and standard token baselines. Controlled development ablations show that specialized supervision contributes beyond standard task fine-tuning and encoder scaling. On GeoNames, the adapted model improves known-target Recall@1 throughout zero-to-moderate character overlap, while character n-grams retain a small aggregate Target Recall@5 advantage. In a blinded human comparison on a stratified production challenge set, our model raises relevant P@1 from 28.0% to 46.0% (p=0.012). Fixed-query endpoint estimates improve on non-canonical queries and remain close to control on frequent queries; a randomized live experiment detects no engagement regression. These results support task-adapted geographic entity retrieval as a practical replacement for the incumbent taxonomy-based standardizer, with the largest relevance gains on non-canonical queries. --- ### RENSA: Rich Environment Metadata to Navigate Shared and Distributed Endpoints for Automated Federated SPARQL Query Generation Authors: Victor Eiti Yamamoto, Takeda Hideaki, Yamamoto Yasunori Published: 2026-08-29 Categories: cs.DB, cs.CL arXiv: https://arxiv.org/abs/2608.28963v1 The number of knowledge graph databases has increased significantly with the proliferation of knowledge graph technologies. Knowledge graphs enable the dynamic integration of distributed data through federated SPARQL queries. However, constructing efficient queries in a federated environment is challenging due to the lack of detailed structural knowledge across decentralized datasets. While standards like VoID provide basic metadata, they often fail to capture the complex interlinks and authority distributions necessary for optimization. Consequently, current engines frequently rely on runtime ASK queries for source selection, increasing communication overhead. We propose RENSA, a federated SPARQL query generation framework that leverages an extension of SPARQL Builder Metadata (SBM). By integrating class and authority information, mapping subject and object usage to specific predicates, RENSA enables precise source selection and semantic constraint inference for query variables without runtime communication. The generated profiles represent less than 1\% of the original dataset triples in most cases, ensuring storage efficiency. Evaluation on the LargeRDFBench benchmark (13 datasets with >1B triples, 32 queries) shows that RENSA achieves source selection results comparable to state-of-the-art methods while eliminating ASK query overhead. Furthermore, we demonstrate that RENSA infers class and authority constraints for query variables, enabling the identification of data sources even across heterogeneous endpoints. These profiles additionally offer human-readable structural insights for semi-automated query generation. --- ### Continuity-Free Near-Minimax Leading-Order Regret for CVaR-UCBVI Authors: Yuanlong Chen Published: 2026-08-29 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.28960v1 For finite-horizon tabular CVaR reinforcement learning, prior work proves a $\widetilde{O}(τ^{-1}\sqrt{SAK})$ leading regret bound for arbitrary normalized return laws and the sharper $\widetilde{O}(\sqrt{SAK/τ})$ rate under a density lower bound. We show that the same Bernstein CVaR-UCBVI algorithm attains the sharper rate without continuity assumptions. The key is a selected-budget self-bound: the conditional variance of the episode shortfall is at most $τ$ plus the value-estimation width. Substitution into the original Bernstein decomposition yields, with high probability, $\widetilde{O}(\sqrt{SAK/τ}+(SAHK^{1/4}+S^2AH)/τ)$ regret for arbitrary normalized return laws, including atomic, mixed, and continuous laws. The $τ^{-1/2}$ leading term matches the expected-regret minimax lower bound up to logarithmic factors. Thus Bernstein CVaR-UCBVI is minimax-optimal over the full return-law class in the leading-order regime; the lower-order terms retain their $τ^{-1}$ dependence. --- ### CoVA-SFT: A Large-Scale Dataset for Chain of Visual Abstractions Authors: Tsung-Han Wu, Heekyung Lee, Anya Ji, Haoming Chen, Trevor Darrell, Joseph E. Gonzalez, David M. Chan Published: 2026-08-29 Categories: cs.CL, cs.CV arXiv: https://arxiv.org/abs/2608.28958v1 Chain-of-thought (CoT) reasoning has dramatically improved large language models (LLMs) by allowing them to decompose problems into intermediate steps. While CoT is widely effective for linguistic tasks, text-only CoT forces models to serialize visual problems into awkward prose. Although architectural solutions exist to process visual inputs, the community lacks a massive, multi-step, self-corrected dataset to teach models how to build and maintain internal visual workspaces when solving purely textual reasoning problems. To address this limitation, we introduce CoVA-SFT, a highly structured corpus of 51.9K samples containing over 222K multimodal reasoning steps across 5 distinct layout families and 17 complex tasks, and CoVA-Bench, a companion benchmark of 1,700 held-out test samples spanning the same tasks for reproducible evaluation. By providing explicit rationale formulations, agentic renderings, and verification loops, CoVA-SFT teaches multimodal language models to interleave text and visual abstractions. We validate the dataset by demonstrating that models fine-tuned on CoVA-SFT outperform all interleaved CoT baselines by more than 2x on average on CoVA-Bench, though they still fall short of strong text-only CoT baselines, highlighting open challenges for future work. --- ### Operator Packages, Proposer Strength, and Construction-Family Plateaus in Office-Scale Verified Search Authors: Roberto I. Ono Filho Published: 2026-08-28 Categories: cs.CL, cs.AI, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2609.29636v1 Verified search, in which a language model proposes programs, a hard evaluator scores them, and selection keeps the best, has recently moved mathematical records; controlled ablations of the proposer-side components remain rare. We instrument a minimal FunSearch-style loop at office scale (a 30B local model on a laptop, 120-600 verified samples per run) with three operator packages: a schematic notebook the model writes and carries instead of verbatim elites, a named obstacle, and behavioural repulsion from constructions already found. On nine construction problems from a public repository, the complete 2^3 factorial with two replicates favours the primary contrast in a nominal two-stage analysis: the composition closes more of the seed-to-record gap (+0.196; nominal pooled p=0.023, stage-combination p~0.08; median per-problem effect +0.045). Repulsion raises construction-hash diversity everywhere (p=0.0039; partly a manipulation check). The factorial finds no positive memory-by-repulsion interaction (bounded to about +/-0.04); the gain decomposes additively, and memory+repulsion is the only arm that never collapses (0 of 18 runs), within 0.025 of the full composition. A frontier proposer under the identical loop reaches in tens of samples what the local model does not in hundreds; in single scoping runs its gains arrive without the operators. The search stalls after closing ~92% of the gap on the flagship problem, and the registered family-hint test gives the stall its first reading: named in words, the reference family is adopted and loses; handed as code, it is optimized, but our best finite-grid implementation remains below the plateau reached unaided. The loop transported and optimized the idea it was handed; no unaided run produced it. We release the harness, every candidate, and the dated pre-registrations. --- ### iCoder-27B: Recursive AI-Led Development of Frontier Industrial Coding Model Authors: Cheng Yang, Jiayang Lyu, Shangyuan Liu, Guibin Zhang, Jiong Lin, Xinlei Yu, Junchi Yan, Shuicheng Yan, Weinan E, Linfeng Zhang, Linfeng Zhang, Qibing Ren Published: 2026-08-28 Categories: cs.AI, cs.CL, cs.SE arXiv: https://arxiv.org/abs/2609.29626v1 Recursive AI, the prospect of AI taking an increasingly complete role in building and improving AI, is a crown jewel of AI for AI. Although recursive self-development has become practical for small models, bounded tasks, and fixed time budgets, a more consequential realization of this ambition, i.e., developing a release-ready, frontier-competitive model, remains far more challenging. In this work, we ask how little human involvement is sufficient for an agent to develop a frontier model. We concentrate human input into a high-density, low-frequency interface: experts encode objectives, stage scaffolds, permission boundaries, and operating procedures as reusable research skills, while the agent instantiates these priors, selects experiments, diagnoses outcomes, and revises the training strategy. In the challenging domain of industrial coding, the agent evolves data and coordinates SFT, on-policy self-distillation, and reinforcement learning with verifiable rewards, ultimately producing iCoder, a 27B model for RTL design and GPU kernel optimization. Across seven benchmarks, iCoder leads RTLLM, outperforming GPT-5.5 and Claude-Opus-4.8; ranks second on CVDP and KernelBench L2, exceeding GPT-5.5 by 16 points; and ties Claude-Opus-4.8 for the best TritonBench result. Exploratory case studies further show iCoder's competitive iterative RTL and GPU-kernel optimization with substantially fewer tokens. These results chart an engineering path toward recursive self-improvement, in which humans distill the principles of model building, agents operationalize them through evidence-driven experimentation, and each generation of AI becomes a more capable architect of the next. --- ### Limited Structural Reliability in Public Educational Prediction Benchmarks: A Four-Dimension Audit of Seven Datasets Authors: Yan Ma, Lizhuo Zhang Published: 2026-08-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2609.29625v1 Across seven public educational prediction datasets, three passed all four pre-modeling reliability checks; the remaining four either failed group-aware generalization tests or lacked the provenance metadata needed to run them. One dataset was initially classified as failing but corrected after excluding group-identifier features from the holdout matrix, demonstrating that the audit can distinguish genuine cross-group confounding from feature-encoding artifacts. Each dataset was audited before model optimization using four checks: baseline gap, split instability, null separation, and metadata adequacy under group-aware holdout. The dominant failure mode was not weak iid performance alone but cross-group fragility: in the clearest case, UCI Student declined from iid R-squared 0.242 to group-holdout R-squared -0.097, while Higher Ed collapsed from 0.041 to -8.79. Increasing model complexity did not remove this pattern: ensemble models improved structurally sound datasets but amplified instability or failed under group holdout on fragile ones. An exploratory cross-dataset comparison further showed that stronger profiles clustered in larger, richer-grouped, performance-proximal datasets, while random-split performance severely overstated deployable signal in fragile datasets. Classification-metric sensitivity analyses reached the same substantive conclusions. The results show that benchmark reliability in educational AI is constrained less by algorithm choice than by data structure, group heterogeneity, and evaluation design. A reusable pre-modeling audit offers a minimum quality gate before public educational datasets support strong benchmark or deployment claims. --- ### AgenticCADedit: A Stateful, Tool-Mediated Agentic Approach to Multimodal 3D CAD Editing Authors: Saptarshi Neil Sinha, Mika Silvan Goschke, Paul Julius Kühn, Arjan Kuijper, Michael Weinmann Published: 2026-08-28 Categories: cs.CV, cs.GR arXiv: https://arxiv.org/abs/2609.29621v1 Computer-aided design is central to industrial manufacturing, and much of a designer's daily work consists of editing existing models from multimodal requests involving speech, sketches, and model interaction. Existing neural CAD approaches focus predominantly on unconditional or text-conditioned generation. The neuralCAD-Edit approach formalizes expert multimodal editing requests, but its iterative baseline refines a complete CAD program across attempts, executing each attempt from the original model in a stateless CAD environment. Every attempt must therefore reconstruct the entire edit from scratch, so partially correct progress is discarded rather than accumulated, and the model can neither inspect the geometry it has just produced nor selectively revert a single faulty operation. We present AgenticCADedit, which turns editing into a sequence of small, verifiable actions on a persistent CAD state instead of a single regenerated program. Rather than emitting one complete program, it applies incremental code steps that each commit to the session, inspects the resulting faces and edges, renders highlighted selections to verify that the intended region was addressed, and reverts individual operations when it was not. Subsequent actions therefore build on the geometry produced by earlier ones. Our approach improves on all metrics for all three evaluated LLMs (open-weight: qwen3.6-27b, gemma4-31b; proprietary: gpt-5.6-luna), with the largest gains for the weakest baseline model, qwen3.6-27b, whose validity rises from 51.0% to 94.8% and acceptance from 1.6% to 12.0%. A token-cost analysis with gpt-5.6-luna further shows $66.7$% fewer output tokens than neuralCAD-Edit, while $94.8$% of input tokens are served from the prompt cache. --- ### An Exploratory Ablation of a Small MLA--SSM Hybrid Language Model Authors: Christos Koutsiaris Published: 2026-08-28 Categories: cs.CL, cs.AI, cs.IR, cs.LG arXiv: https://arxiv.org/abs/2609.29618v1 We report an exploratory, single-seed ablation of TALH (Adaptive Latent Hybrid), a decoder-only language model with parallel Multi-head Latent Attention (MLA) and a custom recurrent state-space (SSM) branch. Five variants, spanning 117--217M estimated active parameters per token, are trained from scratch on a FineWeb sample for the same number of optimisation steps and tokens. In this specific setup, removing the SSM branch gives the largest degradation in validation perplexity (MLA-only PPL 315), whereas removing MLA has a much smaller effect (SSM-only PPL 239). A dense-FFN hybrid obtains PPL 231, compared with 240 for the tested top-2 ternary-MoE hybrid, while using 3.87 GB less peak training memory. We also preserve a preliminary Apple M3 timing observation: among the five unoptimised implementations, MLA-only has the flattest measured time-to-first-token curve from 512 to 2,048 prompt tokens, although the dense Transformer is much faster in absolute terms. Because the runs are single-seed, parameter counts are unmatched, the evaluation stream may overlap the training source, and raw repeated timing records are unavailable, these results support implementation-specific hypotheses rather than general conclusions about MLA, SSMs, or mixture-of-experts models. --- ### Evidence-Driven Differential Diagnosis of Malignant Melanoma Authors: Naren Akash, Anirudh Kaushik, Jayanthi Sivaswamy Published: 2026-08-28 Categories: eess.IV, cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2609.29613v1 We present a modular and multi-level framework for the differential diagnosis of malignant melanoma. Our framework integrates contextual information and evidence at the lesion, patient, and population levels, enabling decision-making at each level. We introduce an anatomic-site aware masked transformer, which effectively models the patient context by considering all lesions in a patient, which can be variable in count, and their site of incidence. Additionally, we incorporate patient metadata via learnable demographics embeddings to capture population statistics. Through extensive experiments, we explore the influence of specific information on the decision-making process and examine the tradeoff in metrics when considering different types of information. Validation results using the SIIM-ISIC 2020 dataset indicate including the lesion context with location and metadata improves specificity by 17.15% and 7.14%, respectively, while enhancing balanced accuracy. The code is available at https://github.com/narenakash/meldd. --- ### STRAND: Benchmarking and Improving Object-Centric Spatio-Temporal Monitoring in Video Large Language Models Authors: Thong Nguyen, Tri Cao, Khoi Le, Cong-Duy Nguyen, Quynh Vo, See-Kiong Ng, Bryan Hooi Kuen-Yew Published: 2026-08-28 Categories: cs.CV, cs.CL arXiv: https://arxiv.org/abs/2609.29607v1 While multimodal large language models (MLLMs) have advanced video understanding, they remain highly prone to hallucinations in dynamic scenes. We argue this stems from a failure in spatio-temporal monitoring, the ability to persistently track object identities, states, and relations over time. Existing benchmarks obscure this deficit by relying on single final-answer evaluations for queries that can often be resolved via local visual cues or statistical priors. To rigorously diagnose this, we introduce STRAND, a benchmark of human-verified object-centric facts that evaluates intermediate reasoning by decomposing queries into sub-questions, distinguishing genuine temporal understanding from coincidental correctness. Crucially, we score models with Faithful Accuracy, an unconditional joint metric that credits a prediction only when the target answer and every prerequisite sub-question are correct, so that a model cannot inflate its score by being selectively consistent on the small subset of targets it happens to answer correctly. To address failure modes exposed by STRAND, we further propose an object-centric framework that explicitly constructs and reasons over structured object trajectories via chunk-wise state extraction and temporal aggregation. Extensive experiments, including backbone-, frame-, call-, and token-matched comparisons against both end-to-end MLLMs and modular video harnesses, demonstrate that our object-centric framework significantly reduces hallucinated answers and improves spatio-temporal reasoning consistency over state-of-the-art MLLMs. The code, model, and data have been made available at nguyentthong.github.io/strand. --- ### PROVE: Proof-guided Regime-aware Operator Verification for Hallucination Detection in Medical Visual Question Answering Authors: Keyang Zhou, Siyi Li, Zhongnan Shi, Qichao Ying, Wei Tang, Zhenxing Qian Published: 2026-08-28 Categories: cs.CV, cs.MM arXiv: https://arxiv.org/abs/2609.29604v1 In medical visual question answering (VQA), hallucinations of vision-language models (VLMs) may lead to confident but incorrect responses, raising the risk of diagnostic errors. Existing hallucination detection methods uniformly estimate the reliability of VLM outputs from response consistency or visual evidence. However, such uniform verification across questions ignores question-specific characteristics, resulting in missed overconfident errors and false alarms from over-verification. We present PROVE (Proof-guided Regime-aware Operator Verification), a black-box detector that adapts verification strategy to the evidential structure of each question. PROVE classifies questions into three verification regimes based on what kind of visual proof they demand, activates a regime-specific subset of five complementary operators, and adjusts operator importance per question through a lightweight calibration layer conditioned on deterministic question-answer features. PROVE uses question-specific evidence to reweight operators and produce a calibrated risk score. Evaluated on 8048 test samples across three medical VQA benchmarks and four frontier VLMs, PROVE achieves 0.821 AUROC, outperforming the strongest baseline by +0.159, with consistent gains across all models and benchmarks. --- ### When Agents Act Unwatched: The Reduced-Supervision Paradox in Agentic AI Authors: Hanjing Shi, Dominic DiFranzo Published: 2026-08-28 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2609.29547v1 Agentic AI is sold on a simple promise: the system keeps acting when the user stops watching. That promise creates an accountability inversion. As stepwise supervision recedes, verification does not disappear; it moves into the runtime infrastructure that defines authority, records action, interrupts execution, checks outcomes, and supports repair. We call this the reduced-supervision paradox. Using a 63-artifact audit, we examine its public visibility across 46 research papers and 17 engineering, documentation, security, and governance sources. We find that agents' action surfaces are far easier to reconstruct than the mechanisms needed to answer for their actions. Tool mediation and monitoring traces were clearly visible in 40 and 37 artifacts, whereas checkpoint placement was clearly visible in 6, validator independence in 4, recovery in 2, and contestability in 1. Three action paths show why this imbalance matters. A repository path can preserve rich diffs after a consequential change. A browser path can cross organizational boundaries faster than permissions travel. A service path can follow policy while leaving affected people without recourse. We argue that observability can become a substitute for accountability when it shifts verification onto users after meaningful intervention is no longer possible. Our action-path diagnostic instead asks whether a delegated action remains connected to authority, evidence, interruption, independent judgment, recovery, and challenge. The claim is deliberately bounded to public visibility; it does not establish the prevalence or effectiveness of undisclosed controls. We contribute an action-level account that recasts the harness from a technical wrapper into accountability infrastructure. --- ### Physics-Informed Classical and Quantum Neural Networks for One-Dimensional Schrodinger Eigenvalue Problems Authors: Tariq Mahmood, Waqas Arshad, Bilal Naseer, Alfredo Raya Published: 2026-08-28 Categories: quant-ph, cs.LG, hep-ph, hep-th arXiv: https://arxiv.org/abs/2609.22189v1 The Schrodinger equation in one spatial dimension admits a small set of exactly solvable potentials that serve as natural proving grounds for any new eigenvalue solver. We formulate Physics-Informed Neural Networks (PINNs) and Physics-Informed Quantum Neural Networks (PIQNNs) for the time-independent Schrodinger equation and apply them to three of these benchmarks: the harmonic oscillator, the infinite square well, and the finite square well. In each case a composite loss encodes the differential-equation residual, the normalization condition, the boundary behavior, and the orthogonality between eigenstates, so that the trial wave function is driven toward a genuine eigenfunction without supervised data. The eigenvalues and wave functions returned by both methods are compared against the exact spectra and against three classical references: the matrix Numerov method, the finite difference method, and the shooting method. For the smooth oscillator the two neural solvers reproduce the lowest four eigenvalues to parts per million, while for the square wells they recover the analytic levels with comparable fidelity even where the potential is discontinuous. The quantum circuit, built as a layered angle-embedding ansatz with strongly entangling blocks, converges more reliably than its classical counterpart on the higher excited states, where the loss landscape of the classical network becomes harder to navigate. --- ### Fairness Beyond Anonymization? Demographic Leakage in German LLM-Generated Resumes Authors: Charlotte Leininger, Helena Veit, Matthias Aßenmacher, Andreas Bender Published: 2026-08-28 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2609.22188v1 Large language models (LLMs) are increasingly integrated into AI-assisted hiring pipelines, including automated resume generation and screening. Under the EU AI Act, the hiring domain is classified as high-risk, making fairness and transparency critical requirements. Existing work has primarily focused on explicit hiring decisions, while less attention has been paid to whether generated resumes themselves encode recoverable demographic information. In this work, we conduct a two-stage audit of demographic leakage in German-language LLM-generated resumes. First, we use ChatGPT (GPT-4o-mini), Gemini 2.5 Flash-Lite, and multiple scales of the open-weight Qwen 3 model family (4B, 8B, and 14B) to generate resumes from real anonymized job-matching profiles, systematically varying gender- and ethnicity-associated names while holding qualifications constant. Second, we simulate a downstream resume screening scenario, where the generated resumes are first anonymized and gender-neutralized, before demographic leakage classifiers are trained on the resulting texts. We find that, despite these interventions, classifiers reliably distinguish between resumes generated with male and female names. This leakage is not driven by overtly gendered wording, but by subtle differences in the usage of semantically equivalent, formally gender-neutral terms in German. In contrast, ethnicity-related leakage remains comparatively weak across models. Our findings demonstrate that apparently neutral resume generation can still preserve highly predictive demographic signals, raising concerns about anonymization-based fairness interventions in multilingual AI hiring pipelines. --- ### Correlation-Guided Flow Matching with Annealed Masking for Spatial Transcriptomics Generation Authors: Yupei Zhang, Hao Chen, Li Pan, Chao Li, Xiaohan Xing Published: 2026-08-28 Categories: cs.LG, cs.CV, q-bio.GN, q-bio.QM arXiv: https://arxiv.org/abs/2609.22187v1 Spatial transcriptomics (ST) provides spatially resolved gene expression profiling but remains expensive, motivating the prediction of ST from histology images. Generative models have emerged as a mainstream paradigm for ST prediction due to their ability to model the conditional distribution of gene expression and capture its inherent stochasticity. However, these methods typically treat genes as independent prediction targets and overlook the intrinsic gene-gene interactions in biological systems, which limits their ability to preserve biologically meaningful co-expression patterns. We argue that gene-gene interactions, which reflect shared pathways and regulatory mechanisms, are essential for generating numerically accurate and biologically coherent ST profiles. In this paper, we propose CorrFlow, a correlation-guided flow matching framework for histology-to-ST prediction that explicitly models gene-gene dependencies through two complementary mechanisms. First, we introduce an annealed masked flow matching strategy, where subsets of genes are progressively masked following a timestep-dependent annealing schedule, encouraging the model to infer masked genes conditioned on the remaining genes and promoting joint conditional modeling beyond per-gene marginal estimation. Second, we devise a gene graph-regularized optimization scheme that integrates prior knowledge from the STRING database and data-driven co-expression estimated by WGCNA to construct a gene affinity graph, which enforces both local consistency and global smoothness in the predicted expression. Extensive experiments across 12 datasets show that CorrFlow achieves the best average PCC and HPCC among evaluated methods, leading to more biologically coherent ST predictions. --- ### Adaptive Physics-Informed Neural Networks for the Blasius Boundary-Layer Problem Authors: Mehari Fentahun Endalew, Xiaoming John Zhang Published: 2026-08-28 Categories: cs.LG, physics.flu-dyn arXiv: https://arxiv.org/abs/2609.22185v1 Physics-informed neural networks (PINNs) provide a mesh-free approach for solving differential equations, but their performance can depend strongly on loss weighting, collocation placement, and optimization strategy. This study develops an adaptive PINN framework for the Blasius boundary-layer equation using gradient-norm-based adaptive loss weighting, nonuniform and residual-based collocation, and sequential Adam--L-BFGS optimization. In the representative run using the architecture $(1,100,100,1)$, the model predicts $f''(0)=0.3320762918$, compared with the high-accuracy benchmark $0.332057336215$, giving an absolute error of $1.896\times10^{-5}$. The final weighted loss is $6.789\times10^{-8}$, and the predicted stream-function, velocity, and shear profiles agree closely with an independent numerical boundary-value solution. A separate full-training architecture study shows that the two-hidden-layer model achieves the smallest wall-shear error among the four tested architectures, $1.629\times10^{-6}$, whereas the deepest network attains the smallest weighted objective but a substantially larger wall-shear error. Compared with the previously reported PINN value $f''(0)=0.33165$, the representative run reduces the wall-shear error by approximately a factor of $21.5$. The results show that the combined adaptive training framework can achieve high accuracy for the Blasius problem and that weighted loss alone is insufficient for identifying the most physically accurate PINN. Because the adaptive components are applied jointly, their individual contributions cannot be isolated from the present results and would require a controlled ablation study for separate assessment. --- ### DPTM-DT: Dual-Pretrained Transformer Multitask Representation Learning for Drug-Target Prediction Authors: Ge Kong Published: 2026-08-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2609.22184v1 Drug-target relation prediction supports candidate screening, drug repositioning, and mechanism analysis. Existing models often use incomplete drug or protein representations, model cross-modal interactions shallowly, or train affinity regression and interaction classification separately, although these tasks describe closely related views of the same drug-target pair. This paper presents DPTM-DT, a dual-pretrained Transformer framework for multitask drug-target prediction. DPTM-DT combines GROVER molecular graph embeddings, ESM protein language-model embeddings, and CTD physicochemical descriptors, then exchanges drug-target information through bidirectional cross-modal attention. A shared pair representation is used for continuous affinity regression, high-affinity binary classification, and six-level affinity classification. Experiments on Davis and KIBA cover random 80/20 and DeepDTA-style standard splits. On the random 80/20 split, DPTM-DT achieves MSE/CI values of 0.193/0.917 on Davis and 0.120/0.918 on KIBA. It also reports binary AUPR/MCC values of 0.727/0.654 and 0.798/0.689, and six-class Macro-F1/Top-2 values of 0.800/0.932 and 0.815/0.962 on Davis and KIBA, respectively. Across the reported regression, binary classification, and multiclass classification settings, DPTM-DT achieves the best overall performance among the compared methods. Results under the standard split show the same relative trend. Ablations indicate that dual target representation, gated fusion, and cross-modal attention each contribute to the final performance. Code and supplementary materials are available at: anonymous.4open.science/r/DPCM-DT-74E0. --- ### Generic Characteristic-Zero Equivalence Between Derivative Bézout Inversion and Multipoint Evaluation Authors: Zijian Zeng Published: 2026-08-28 Categories: cs.SC, cs.LG arXiv: https://arxiv.org/abs/2609.17578v1 Let $a_1,\ldots,a_m$ be distinct elements of a field $K$, and let $Z(X)=\prod_{i=1}^m (X-a_i)$. We study the arithmetic complexity of computing the unique normalized Bezout pair $s,t$ satisfying $sZ+tZ'=1$, with $°t<m$ and $°s<m-1$. The classical product-tree approach requires $O(M_K(m)\log m)$ field operations, where $M_K(m)$ denotes the cost of multiplying degree-$<m$ polynomials over $K$. Thus, even when $M_K(m)=O(m\log m)$, the resulting bound is $O(m\log^2 m)$ rather than $O(m\log m)$. Over an infinite field of characteristic zero, we prove that, in the generic rational straight-line-program model, computing all coefficients of the canonical Bezout pair is equivalent, up to an additive $O(M_K(m))$ cost, to arbitrary-node multipoint polynomial evaluation and to interpolation. The main ingredient is an explicit differential reconstruction that recovers $Z$ from $(s,t)$ in $O(M_K(m))$ arithmetic operations on a nonempty Zariski-open subset. Combining this reconstruction with automatic differentiation and transposition yields the complexity equivalence. We further transfer Strassen's lower bound for the elementary symmetric functions to the Bezout problem, obtaining an $Ω(m\log m)$ nonscalar lower bound. Hence an $O(m\log m)$ algorithm, if it exists in this model, would be asymptotically optimal. The reconstruction is genuinely characteristic-dependent: in characteristic $p$, all squarefree polynomials $X^p+cX+d$, with fixed $c\ne 0$, have the same normalized Bezout pair $(0,c^{-1})$. Therefore the all-field, all-input $O(m\log m)$ problem remains open; generically in characteristic zero, however, it is reduced to the corresponding arbitrary-node multipoint evaluation and interpolation problem. --- ### A Survey on Adversarial Attacks and Defenses for Diffusion Models Across Multiple Modalities Authors: Ozgur Kara, Tarik Can Ozden, Furkan Horoz, Zeqian Long, Haotian Xue, Yipu Chen, Oguzhan Akcin, Yongxin Chen, James Matthew Rehg Published: 2026-08-28 Categories: cs.CV, cs.CR arXiv: https://arxiv.org/abs/2609.05503v1 Diffusion models have become the dominant family of generative models in the visual domain. However, their widespread public availability enables misuse at scale, motivating a rapidly growing body of research on adversarial attacks and defenses. This survey provides, to our knowledge, the first unified review of this literature across three visual modalities: image, video, and 3D. We introduce a comprehensive, task-centric taxonomy: we first divide the literature by modality; within each modality, we separate methods into attacks and defenses, and then group them by the generative task they target, presenting them chronologically within each task. Moreover, we provide an in-depth analysis of their evaluation settings, consolidating the datasets, metrics, and benchmarks used to assess them. We conclude by identifying several open challenges and outlining concrete future research directions. Project Webpage: https://github.com/ozgurkara99/awesome-adv-attack-defense-on-diffusion --- ### Motion-Omni: End-to-End Joint Speech and Full-Body Motion for Spoken Dialogue Authors: Chengqian Ma, Wei Tao, Haoyu Zhang, Yiwen Guo Published: 2026-08-28 Categories: cs.SD, cs.CV, eess.AS arXiv: https://arxiv.org/abs/2609.04250v2 An avatar that holds a conversation should decide what to say and to move while saying it, yet these abilities live in separate model families: spoken dialogue models produce speech without motion, and co-speech motion models produce motion only from audio handed to them. The standard remedy is a cascade that first generates the spoken response and then runs a motion model over the finished audio, which requires a second full inference pass and precludes any joint optimisation between the two. We present Motion-Omni, an end-to-end framework in which a spoken dialogue model natively outputs explicit facial expression together with hand, upper-body and lower-body motion, generated directly from the hidden states that produce the speech. Joint training is not optional here: with the speech pathway frozen, motion remains misaligned with the audio, and co-adapting the LLM, Speech Generator and Motion Generator under both objectives is what recovers alignment while retaining spoken-dialogue ability. Supervision comes from a scalable, model-agnostic pipeline that pseudo-labels consistent-voice speech responses with a replaceable motion teacher, yielding 422,856 quality-ranked pairs (1,402 hours). We further release SwDA-500 and, to our knowledge, the first public evaluation protocol for stochastic open-ended full-body spoken dialogue, matching audio across motion systems while unifying rendering, automatic metrics, human evaluation, and latency measurement. Instantiated with a Qwen2.5-7B-Instruct backbone, Motion-Omni-Q7 matches the same-audio teacher cascade to within 2% on reference-free motion metrics while responding 5.4 x faster (RTF=0.78, faster than real time), surpasses all non-teacher cascades on beat correlation and diversity, and reaches a 2.62% word error rate, the lowest among the omni-modal systems compared. --- ### Encore: Infinite Audio-Video Generation with Adaptive Signal Routing Authors: Shaohua Pan, Junbao Chen, Shengyi He, Jingfeng Xue, Wen Tao, Haocheng Feng, Siming Fan, Dongwei Pan, Yi Yang, Wei He, Hang Zhou Published: 2026-08-28 Categories: cs.MM, cs.CV, cs.SD arXiv: https://arxiv.org/abs/2609.04249v1 Existing audio-video generation methods produce well-synchronized clips but are limited to short durations, while long-video generation methods extend duration through chunk-based iterative synthesis yet lack audio entirely. Generating long audio-video jointly is fundamentally harder than either task alone: each chunk must simultaneously maintain video temporal coherence, audio temporal coherence, and cross-modal synchronization, whose conditioning signals enter the model through different pathways. In this work, we present Encore for long-form synchronized audio-video generation. Our key insight is to factor this challenge into: (1) local continuity handled by iterative generation with explicit cross-chunk context propagation, and (2) global consistency enforced via reference audio-video signals with shifted position embedding. Building on this design, we propose Adaptive Signal Routing (ASR), which introduces learnable attention biases within self-attention and learnable residual scales on cross-attention outputs, enabling the model to adaptively modulate the influence of each conditioning signal. Trained end-to-end for joint audio-video generation, Encore also supports infinite-length audio-to-video and video-to-audio synthesis at inference by conditioning on the ground-truth modality throughout the denoising process. Experiments on our extended VerseBench for long audio-video evaluation demonstrate that Encore significantly outperforms existing methods in both generation quality and temporal coherence. Code and data for this paper are at https://github.com/shaohua-pan/Encore. --- ### SocialBuddy: Tailoring Search Agent for Social Scenarios Authors: Mingxuan Li, Yirong Mao, FaZhan Zhang, Haibiao Yao, Runze Hu, Wenhui Que Published: 2026-08-28 Categories: cs.SI, cs.LG arXiv: https://arxiv.org/abs/2609.01641v2 In the era of digital social interaction, searching friends' posts from massive social streams has become a fundamental user need. However, while modern agentic search frameworks have achieved remarkable success in conventional retrieval tasks, they break down when confronted with heterogeneous user queries and multi-dimensional social feeds, resulting in severe performance degradation in complex social search. To bridge this gap, we introduce SocialBuddy, the first agentic search framework tailored for social scenarios. Specifically, we construct SocialEnv, the first large-scale simulated environment for social search. Powered by an automated data and trajectory synthesis pipeline, SocialEnv includes 200K user profiles, 10 million social posts, and 50K reasoning trajectories, establishing a solid foundation for the development of social search agents. To tackle the credit assignment dilemma caused by sparse rewards in social search, we design SocialPO, a hybrid-granularity optimization framework. It macroscopically reinforces successful reasoning paths via multi-dimensional rewards, while microscopically rectifying deviated trajectories through fine-grained prefix truncation and token-level supervision. This hybrid-granularity design delivers multi-scale guidance in complex long-sequence scenarios. Finally, we construct SocialSearch Benchmark to provide a quantitative evaluation scheme for assessing the social search capabilities of SocialBuddy. Extensive experiments demonstrate that SocialBuddy-35B surpasses significantly larger frontier LLMs. Code and dataset will be released upon article acceptance. --- ### EULER: Exploring Underused Links with Evidence-Checked Return for Multi-Agent Mathematical Discovery Authors: Ren Zhenzhuo Published: 2026-08-28 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2609.00032v1 Mathematical communities work with different objects, invariants, and tools, so transferring a problem across them is expensive and often skipped. We present EULER, a multi-agent system that takes such a transfer--a bridge--as its unit of search. Around a fixed conjecture, EULER runs direct, adjacent-domain, and distant-domain routes in competition; a bridge keeps its budget only if it supplies an operation the source representation cannot execute and its target-side evidence returns to the original statement along a checked implication. Six ordered stress tests reject invalid bridges before expensive search begins. We evaluate EULER on 120 recent conjectures. The conjectures were frozen before search and screened for contamination, and are drawn from public papers by authors who had recently published in the Journal of Combinatorial Theory, Series A, a leading journal in combinatorics. EULER produced 10 proofs and 3 refutations, plus 45 scoped partial results. Two mechanisms held up under ablation: bridge-specific stress tests cut incorrect conclusions from 9 to 3, and bridge material combined with a target-native operation yielded a positive interaction of +4.2 resolved tasks that neither factor produced alone. Domain distance did not reliably predict success; executable operation gain and valid return did. --- ### The information geometry of product-reference discrete diffusion: Interaction growth complexity and optimal scheduling Authors: Martin J. Wainwright Published: 2026-08-28 Categories: stat.ML, cs.AI, cs.LG, math.ST arXiv: https://arxiv.org/abs/2608.28949v1 We study a class of product-reference diffusion algorithms for sampling from a discrete distribution. We show that their sampling performance can be characterized using a path-based measure of data geometry that we call the interaction growth complexity (IGC). We show that a bivariate IGC kernel gives an exact representation of both the KL discretization error and a simple one-step upper bound. The simpler univariate IGC density can be used to study the effect of stepsize choices on the iteration complexity required to obtain $ε$-accurate samples in KL divergence. Samplers that traverse the path with equi-spaced steps in log-squared-reliability-odds have performance that depends on the aggregate IGC mass, whereas refined choices of stepsizes have a lower complexity depending on a square-root functional. In the fine-grid limit, both of these characterizations become sharp. We also allow general product reference distributions and show that the reference law can substantially reshape the IGC profile and the resulting sampling complexity; in particular, references far from both the uniform and the data marginals can yield dimension-dependent improvements. Finally, the aggregate IGC mass admits bounds in terms of total correlation and dual total correlation, thereby connecting the pathwise geometry to classical measures of multivariate dependence. --- ### From the Loss Landscape to Diverse Feature Learning in Neural Networks Authors: David Aram Yunis Published: 2026-08-28 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2608.28948v1 Over the course of the last decade, neural networks have grown from an academic curiosity to moving the markets of nations. Despite this explosion in both research and deployment, relatively little is understood about how they achieve the solutions they do. This is both scientifically relevant, and pressing for society. When neural networks make decisions across self-driving, construction, law, hiring and health, there have been and will continue to be unintended consequences. However, attempting to generalize the failures of the largest and most important production systems makes for a very difficult task. Yet signs of these failures exist at all scales of neural networks, so we should be able to study a much more tractable setting. All neural networks must undergo an optimization process, called training, to be useful. To a great degree, understanding neural networks is understanding their optimization: through what process and exposure to which data did they arrive at their results. Yet our knowledge on this topic as a field is quite imprecise. In particular, a curious phenomenon called mode connectivity, the ability to connect neural networks in the loss surface, defies explanation entirely. This dissertation elucidates, explains and exploits this special structure in the loss landscape... --- ### Automated Researchers Can Mitigate Well-characterized Alignment Failures Authors: Chen Yueh-Han, Jiaxin Wen, Jan Hendrik Kirchner Published: 2026-08-28 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2608.28945v3 Automating alignment research may accelerate progress toward aligned AI, but whether it does is hard to measure. Luckily, many alignment failures, such as deception, sycophancy, and jailbreaks, are already measurable by public benchmarks. We study whether automated alignment researchers (AARs) can post-train to mitigate alignment failures by proposing training methods and data to simultaneously optimize multiple safety benchmarks, while largely preserving general capability. Across 10 alignment failures, the strongest AAR methods significantly reduce the targeted alignment failures and generalize to a held-out benchmark, multi-turn behavioral audits, and models up to 4.7x larger than the target model. As a human baseline, 28 experienced researchers receive up to eight hours to develop one-shot methods for the same benchmarks, but their methods underperform the best AAR methods. Using human ideas as the AARs' initial research direction does not improve performance, suggesting current AARs may not need guidance from experienced researchers. These results suggest that automating alignment research on well-characterized failures may be practical in the near term. --- ### Oculi: A Conversational Agentic Platform for Automated Credit Risk Analysis Authors: Vennise Ho, Kristian Diana, Sandy Mourad, Milena Pilipovic, Vineel Nagisetty, Hossein Hajimirsadeghi Published: 2026-08-28 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.28944v1 Credit risk analysis in financial institutions traditionally requires analysts to manually write SQL queries, run statistical computations, and build visualization dashboards. This is a time-consuming workflow that limits exploration to familiar segments. We introduce \textbf{Oculi}, a conversational platform that transforms natural language questions into comprehensive credit risk analyses, complete with data queries, statistical testing, and interactive visualizations. Oculi employs a three-layer architecture that separates reasoning (LLM-powered agent), execution (Model Context Protocol tool servers), and presentation (agentic UI), enabling analysts to discover high-risk portfolio segments. Within Oculi, a new segment discovery pipeline is proposed that combines deterministic statistical methods with LLM-guided feature selection, leveraging LLM semantic domain knowledge alongside data-driven metrics to identify meaningful, actionable portfolio segments. Evaluated on a mortgage portfolio with 200+ features, Oculi demonstrates effectiveness in discovering material risk segments previously intractable through manual exploration, reducing time-to-insight significantly while maintaining auditability and statistical rigor. --- ### Revisiting the Provable-Auditable Privacy Gap of DP-SGD Authors: Saloni Modi, Srivi Balaji, Yusong Zhu, Gautam Kamath, Kevin Tian Published: 2026-08-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.28934v1 Differential privacy (DP) has traditionally been used to provide theoretical upper bounds on an algorithm's stability to changing its training data. In modern private machine learning applications, achieving strong tradeoffs between utility and theoretical privacy is challenging, and thus one may optimistically hope that existing theoretical privacy analyses are loose. Recent work on privacy auditing has adopted a dual viewpoint, instead lower bounding the true privacy of an algorithm by constructing empirical distinguishing events. The auditing literature has thus far yielded a pessimistic outlook on the looseness of theoretical privacy bounds for DP-SGD, the de facto private training method in modern ML, as nearly-matching empirical lower bounds have been achieved under various threat models (NHSBTJCT23, AC24, CBP25). In this work, we propose the empirical privacy lower bound of an algorithm as a concrete metric to optimize for, complementary to the theoretical upper bound. We give a lightweight defense framework that generically augments optimization methods in the ML pipeline to have significantly-improved empirical privacy on standard benchmarks. Moreover, we show that our framework comes at no theoretical privacy cost when augmenting DP-SGD, unlike previously-proposed defenses against membership inference attacks. We evaluate our defense against a broad range of audit constructions, models, and datasets to demonstrate its flexibility. --- ### NBS: No Bias Stereo Authors: Vage Taamazyan, Zhuowen Shen, Stefan Hinterstoisser, Alberto Dall'Olio, Agastya Kalra, Aarrushi Shandilya, Xin Li, Wenping Wang, Kartik Venkataraman Published: 2026-08-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.28933v2 Stereo reconstruction is one of the last remaining Computer Vision tasks where all state-of-the-art methods employ a heavy architectural inductive bias. Even though it has been demonstrated that the task can be solved using general-purpose methods, it is widely believed that inductive biases in stereo are strictly necessary for both high-quality results and computational efficiency. We challenge this paradigm. In this paper, we demonstrate that both state-of-the-art accuracy and superior runtime efficiency are achievable with a model completely devoid of architectural inductive biases, relying instead on a simple, end-to-end Vision Transformer. By training on massive synthetic datasets, we show that pure data-driven learning can surpass explicitly engineered geometry. This work proves that explicit inductive biases are no longer a prerequisite for stereo matching, ultimately unlocking true scaling laws for continuous improvement in 3D reconstruction. --- ### VocalAffectBench: Evaluating Vocal Emotion Recognition in AI Audio Models Authors: Luc Debaupte, Tyler Baumgartner, Brandon Tai, Candice Fan, Bill Wang, Yi Zhong Published: 2026-08-28 Categories: cs.CL, cs.SD arXiv: https://arxiv.org/abs/2608.28932v2 Voice products increasingly need affective cues that are present in speech but absent from transcripts. We introduce VocalAffectBench, a public, test-only benchmark for evaluating whether AI audio models can identify expressed vocal emotion from raw audio. The benchmark contains 273 human-recorded English WAV clips from 51 speaker accounts totaling 1.95 hours across seven labels: angry, disgusted, fearful, happy, neutral, sad, and surprised, with 39 clips per class. All baselines are evaluated from audio alone, without transcripts or contextual metadata. Across six released baselines, average accuracy is 35.5%. The strongest baseline, gemini_3_5_flash, reaches 46.5% on the seven-way task, above the 14.3% random baseline but far from robust emotion recognition. A secondary valence-bucket analysis maps labels into positive, neutral, and negative classes, excluding surprised because its valence is ambiguous. Aggregate accuracy under this coarser view is 50.9%. Performance is highly uneven across classes. By recall, neutral is identified most reliably at 75.6% averaged across baselines, while surprised and fearful reach only 10.7% and 15.4%, respectively. These results show that the evaluated baselines can extract some affective signal from speech, but discrete expressed-emotion recognition remains fragile, especially for non-neutral emotions that are often most important in voice agent workflows. --- ### MERIT: Mitigating Exposure Bias in Generative XMC for User-Interest Propensity Modeling Authors: Abhinav Mahajan, Arindam Sarkar, Prakash Mandayam Comar Published: 2026-08-28 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2608.28931v1 Matching users to interest categories at scale is central to personalized shopping, but the task is challenging in large e-commerce platforms, where label spaces continually evolve and user-interest signals are sparse and long-tailed. Autoregressive language models are appealing because their world knowledge and semantic priors over descriptors generalize across extreme label spaces and accommodate multiple valid label assignments. Yet under teacher-forced fine-tuning, inference-time predictions become part of the conditioning context: early errors steer later outputs toward co-occurring labels, over-generating near-correlates and missing unrelated true interests. We present MERIT, a framework for user-interest propensity modeling that mitigates this exposure bias through a self-correction objective. A permutation-invariant multi-target loss over shuffled mixtures of gold and mined hard-negative labels exposes the generator to erroneous prefixes while preserving the efficiency of teacher-forced training. This training objective concentrates supervision at classification positions, yielding propensity-aligned hidden states powering a lightweight scorer for bidirectional retrieval (interests for users and users for interests). On a proprietary e-commerce dataset with 250k+ interest categories, MERIT improves global recall by at least 11.9% and average Hit@k by 6.1%. In production A/B tests, it achieves +0.26% gain in user conversion. --- ### The Hallucination Signal Is a Mean Shift: Why Simple Probes Suffice Authors: Jungseob Lee, Jaehyung Seo, Heuiseok Lim Published: 2026-08-28 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2608.28930v1 Hidden-state probes effectively detect LLM hallucinations, but the geometry of the signal remains poorly characterized, driving increasingly complex probe architectures. Across three 7B-scale models and three datasets in a paired-example paradigm, we find the signal overwhelmingly dominated by a single mean-shift component, and removing this direction collapses detection to chance. Shrinkage linear discriminant analysis closes about 73% of the gap between 1D and full-dimensional classifiers, so apparent architectural complexity largely reflects high-dimensional covariance estimation difficulty rather than exploitable non-linearity. A simple L2-regularized logistic regression (0.952 AUROC) bounds or outperforms twelve controlled architectural alternatives, and our multi-layer aggregation exceeds CLAP cross-layer attention probing under matched paradigm. Because the signal spans a contiguous layer band, LayerMix aggregates it to match oracle-layer performance without oracle access. Our claims characterize the geometry within the controlled paired-example paradigm. Our code is available at https://github.com/js-lee-AI/LayerMix. --- ### Membership is Ownership: A Robust Ownership Verification Framework for Diffusion Models Authors: Feng Jiang, Zuobin Xiong, An Huang, Zhipeng Cai, Yingshu Li Published: 2026-08-28 Categories: cs.CR, cs.CV arXiv: https://arxiv.org/abs/2608.28929v1 Large-scale diffusion models have fueled numerous profitable downstream applications for AI-related businesses, including visual editing and content creation. Meanwhile, due to the huge amount of resource consumption (e.g., computation and high-quality data) during training, such diffusion models are deemed valuable intellectual property (IP) for tech companies like OpenAI and Google. Yet, the IP assets are vulnerable to various unauthorized uses by adversaries seeking to steal models for customized, usually commercial applications. Some existing approaches have explored IP protection for AI models; however, they mostly face structural limitations in common --- using a training-time watermarking by injecting artifacts in the model, which can impose a measurable utility cost and can be weakened by post-hoc fine-tuning. To address these challenges, this work investigates IP protection (i.e., model ownership verification) for diffusion models in a realistic commercial scenario with minimal model utility loss. Specifically, the proposed method builds a framework for model ownership verification, termed ``{Membership is Ownership} (MiO)'', based on a population-level hypothesis test on a private member evidence dataset. MiO verifies ownership using two criteria: model attribution through membership inference and model separation from public references. Both are tested at $p<10^{-6}$. We evaluate MiO on DDIM and Stable Diffusion models without modifying the owner model or its sampling pipeline, and report ROC-AUC and true-positive rates at fixed nominal false-positive targets. Furthermore, MiO stays stable under different post-theft fine-tuning and weight perturbation in adversarial scenarios, reflecting better robustness compared to the watermarking methods. --- ### Leveraging Turn-taking Dynamics for Intent Recognition in Multi-party Conversations Authors: Galo Castillo-López, Alexis Lombard, Gaël de Chalendar, Nasredine Semmar Published: 2026-08-28 Categories: cs.CL, cs.LG arXiv: https://arxiv.org/abs/2608.28926v1 We propose a multi-task learning approach for multi-party dialogue intent recognition that leverages an auxiliary task that models turn-taking dynamics. Specifically, we introduce turn-transition entropy, a self-supervised target computed from the sequence of speaker transitions, which quantifies the predictability of interaction patterns. Experiments on multiple pre-trained models demonstrate that incorporating this auxiliary task improves intent recognition performance, outperforming existing approaches which ignore multi-party interaction dynamics. We find that our proposed continuous target can be learned as a single-task objective, suggesting that it is an actual signal carrying useful information. --- ### Causal Interventions Reveal Typologically Organized Syntactic Mechanisms in Multilingual Language Models Authors: Sasha Boguraev, Toshiki Nakai, Kyle Mahowald, Julius Steuer Published: 2026-08-28 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.28924v1 Linguistic theory has long recognized cross-linguistic syntactic regularities, leading to claims that these similar structures are processed by similar mechanisms. However, this hypothesis has been difficult to test empirically due to our lack of fine-grained, manipulable access of human processing mechanisms. In this work, we take advantage of techniques from mechanistic interpretability to study such a question in multilingual LMs. We first isolate language-internal mechanisms before attempting to transfer them cross-lingually. Across four models and three well-studied constructions (subject--verb number agreement, anaphoric pronoun gender agreement, and filler--gap object extraction) we find consistent cross-lingual mechanism transfer. We further find transfer to be graded, with more transfer between more typologically similar languages. We believe our work provides novel hypotheses about cross-linguistic syntactic structures and multilingual processing, and more broadly shows how the study of language models can help inform linguistic theory. --- ### ActiveAugment: Online Active Learning for Augmentation Selection in Deep Learning Authors: Noah Videcrantz, Mostafa Mehdipour Ghazi Published: 2026-08-28 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2608.28923v1 Data augmentation is a cornerstone of deep learning pipelines, yet existing strategies treat it as a static, model-agnostic preprocessing step, either relying on expensive dataset-specific policy search or applying transformations uniformly at random, regardless of what the model has already learned. We introduce ActiveAugment, a unified framework that treats augmentation selection as an online active learning problem. For each training minibatch, ActiveAugment generates a pool of candidate augmented views and scores each candidate using a combination of the model's predictive uncertainty and the feature discrepancy induced by the augmentation. The augmentation under which the current model is most fragile is selected per sample, and the model is then trained with a joint supervised classification and supervised contrastive objective that enforces intra-class invariance to the selected augmentations while maintaining inter-class separation. We evaluate ActiveAugment on eight benchmark datasets spanning natural and medical imaging, using CNN and transformer architectures across three training regimes (training from scratch, full fine-tuning, and linear probing), and comparing eight active selection strategies for augmentation scoring. ActiveAugment outperforms AutoAugment, RandAugment, and TrivialAugment under controlled augmentation shifts across all domains and budgets, with the most pronounced gains at low labelling budgets. On medical imaging datasets, where data is scarce and domain shift relative to natural-image pretrained models is large, ActiveAugment achieves higher test F1 than all baselines, demonstrating strong cross-domain adaptability. Our analysis reveals that the augmentation selection policy evolves meaningfully during training and that strategy choice has a direct impact on generalisation. Code is available at: https://github.com/noahvide/ActiveAugment. --- ### RankShift: In-Database Detection and Explanation of Categorical Shifts Authors: Omair Shafi Ahmed Published: 2026-08-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.28922v1 A login service can receive its usual number of failed sign-ins while one source grows from 2% to 30% of them. The same pattern appears in system logs when a rare event template becomes common while the message rate stays stable. These events change which categories are active without changing how many events occur. RankShift detects such changes inside the analytical database that stores the data. It compares each window's category shares with a benign reference using a Pearson score whose terms identify the categories responsible for the change. The same query returns the score, calibrated alert, and largest increasing contributions. We evaluate RankShift on HDFS, BGL, and Thunderbird. It matches the count-vector autoencoder within 0.001 AUROC on HDFS (0.999 versus 1.000) and leads on Thunderbird (0.983 versus 0.949). In a controlled fixed-volume experiment, RankShift detects rare-category shifts that are invisible to event-count monitoring, reaching 0.787 AUROC compared with 0.771 for the autoencoder. Across all three corpora, observed false-alarm rates track the requested operating levels. RankShift requires no model training or inference service, and the autoencoders deployed state is 137x larger. --- ### VoiceCodeBench: Evaluating Exact Structured-Token Recovery in Automatic Speech Recognition Authors: Tyler Baumgartner, Brandon Tai, Lisa Kaelin-Martin, Candice Fan, Luc Debaupte, Bill Wang, Yi Zhong Published: 2026-08-28 Categories: cs.CL arXiv: https://arxiv.org/abs/2608.28916v2 Automatic speech recognition is usually evaluated with word error rate (WER), although voice workflows often require exact written values. VoiceCodeBench measures whether transcripts preserve identifiers, paths, commands, and other structured tokens needed by downstream software. It contains 300 human-recorded English workplace segments (5.59 hours, 85 speakers) and 1,482 audited entities across 26 types and eight domains. Under a raw-audio-only protocol, we evaluate 19 batch and streaming systems using WER, Canonical Token/Entity Match (CTEM), and strict segment-level Task Success Rate (TSR). Across systems, WER has little rank agreement with CTEM (Spearman $ρ=-0.28$) or TSR ($ρ=-0.22$). The best CTEM and TSR are 91.8% and 68.7%. Even the strongest system therefore leaves nearly one-third of recordings with an unrecovered critical value. Symbol-, separator-, and boundary-sensitive entities account for most errors. --- ### mmIR: Frequency-Space Inverse Rendering for 3D Millimeter-Wave Radar ADC Synthesis Authors: Adnan Armouti, Yixuan Gao, Rajalakshmi Nandakumar Published: 2026-08-28 Categories: cs.CV, cs.GR, cs.LG, eess.SP arXiv: https://arxiv.org/abs/2608.28913v1 High-resolution 3D radar data is scarce. Commodity mmWave sensors use small antenna arrays that limit angular resolution to several degrees, and existing datasets provide only 2D range-azimuth maps or sparse point clouds rather than raw analog-to-digital converter (ADC) signals. Hardware scaling is expensive, synthetic-aperture scanning is impractical at fleet scale, and learned synthesis methods are bottlenecked by the very data shortage they aim to address. We present mmIR, an open-source differentiable frequency-modulated continuous-wave (FMCW) radar inverse renderer that fits a physics-based forward model to real captures and re-renders from dense virtual apertures to synthesize high-resolution 3D radar data. Because radar resolution is too coarse to recover geometry directly, mmIR performs LiDAR-assisted inverse rendering: using LiDAR-derived meshes as a geometric scaffold, mmIR optimizes per-vertex International Telecommunication Union (ITU) physics materials, vertex normals, and antenna beam patterns through end-to-end automatic differentiation of a phase-coherent multiple-input multiple-output (MIMO) forward model with multi-bounce propagation, polarization, and free-space diffraction. On seven outdoor and six indoor ColoRadar scenes, mmIR achieves 0.914 mean Pearson correlation on range-azimuth maps versus 0.307 for Sionna-RT. Scenes trained on a cascaded imaging radar transfer to a co-located single-chip radar without re-training (0.554 correlation), and dense virtual arrays (100x100 elements) produce single-frame 3D occupancy validated against LiDAR. Project page: https://mmwave-inverse-rendering.github.io/ --- ### SemKV: Semantic Mixed-Precision KV Cache Quantization Guided by the Quality Cliff for Long-Context LLM Inference Authors: Daeha Lee, Do-Hyung Kim, Jae-Hong Kim Published: 2026-08-28 Categories: cs.LG, cs.CL, cs.IT arXiv: https://arxiv.org/abs/2608.28911v1 The key-value (KV) cache is the dominant memory bottleneck of long-context large language model (LLM) inference, growing linearly with context length. We show that uniform KV quantization on a fractional-bit grid does not degrade gracefully: under a prespecified multi-seed statistical protocol, Llama-3.1-8B-Instruct with an affine quantizer is statistically indistinguishable from FP16 KV down to 2.322 code bits/value and collapses at 2.0 bits - a quality cliff in (2.0, 2.322) that reappears in generation-time quantization and multi-turn dialogue and transfers to Mistral-7B. The cliff reframes importance-aware mixed precision: above it, eight model-internal importance indicators are statistically interchangeable, so the benefit of mixing is grid interpolation, reaching average precisions uniform quantization cannot realize. SemKV preserves every token, ranks tokens by a model-internal score, and assigns two adjacent above-cliff precisions, achieving a measured 6.0x storage reduction with no statistically detectable quality difference from full KV (n=900, three seeds), and outperforming FP16 token pruning granted a 1.5x larger memory budget. Replacing the affine base with a distortion-optimized quantizer (TurboQuant-MSE) lowers the cliff in every protocol tested, raising the no-detectable-loss operating point to 7.9x. The recipe: measure the cliff for the target deployment setting, then interpolate above it. --- ### Learning-Theoretic Foundation for General Coded Computing: The Straggler Setting Authors: Parsa Moradi, Behrooz Tahmasebi, Mohammad Ali Maddah-Ali Published: 2026-08-28 Categories: cs.LG arXiv: https://arxiv.org/abs/2608.28910v1 Coded computing has emerged as a powerful paradigm for mitigating the impact of straggling workers in distributed computing systems. However, existing coded-computing schemes are predominantly designed for the exact recovery of highly structured computations, such as polynomial evaluation and matrix multiplication, and typically rely on strict recovery thresholds. These assumptions significantly limit their applicability to modern machine-learning workloads, particularly deep neural networks (DNNs), whose computations generally lack rigid algebraic structure and, in many applications, require only accurate approximations rather than exact recovery. To address this gap, we revisit coded computing from a learning-theoretic perspective and introduce General Coded Computing (GCC). Rather than adopting existing algebraic tools, GCC formulates coded computing through a natural end-to-end mean-squared error loss that directly measures the discrepancy between the desired computations and their recovered estimates. By deriving suitable upper bounds and restricting the encoder and decoder to a reproducing kernel Hilbert space (RKHS) with mild smoothness constraints, we show that both the encoder and decoder admit specific representations as linear combinations of RKHS kernel functions. This representation allows the corresponding coefficients to be computed efficiently. Moreover, this framework enables us to establish theoretical performance guarantees for GCC under two complementary straggler regimes. In the worst-case setting with $N$ worker nodes, and at most $S$ stragglers, we show that the end-to-end loss decays at least at rate $O(S^3N^{-3})$ for standard configurations. We then study a probabilistic setting in which each worker independently straggles with probability $p$. We prove that the expected loss can still converge at rate $O(\log_{1/p}^3(N)N^{-3})$. --- ### Coarse to Fine: Iterative Adversarial Neural Cellular Automata for Medical Image Synthesis Authors: Anh Thi Luu, Nick Lemke, Anirban Mukhopadhyay Published: 2026-08-28 Categories: cs.CV arXiv: https://arxiv.org/abs/2608.28909v1 Large-scale, publicly available datasets have driven advances in deep learning, but privacy and legal restrictions often limit data sharing in medical imaging. Synthetic data generation offers a privacy-friendly alternative to enable the training of high-performance models on health data. While most state-of-the-art generative models produce high-quality images, they remain computationally expensive, which limits their applicability on resource-constrained hardware. We propose StyleGANCA, the first lightweight general-purpose NCA-based generative adversarial network. The architecture integrates a StyleGAN-inspired mapping network and adaptive style modulation into a multi-scale NCA synthesis process, enabling latent-controlled image generation through iterative local interactions. We evaluate StyleGANCA on BloodMNIST and PathMNIST against adversarial, variational, diffusion, and NCA-based baselines. Experimental results demonstrate that StyleGANCA achieves competitive image quality with substantially fewer parameters than baseline architectures, achieving the best FID and KID scores on PathMNIST with only 617k parameters. Furthermore, downstream experiments show that the generated images preserve class-specific information and effectively support the training of multi-class classifiers. Our code is publicly available at: https://github.com/MECLabTUDA/StyleGANCA ---