# Research Papers — 2026-06 · bundle page 202 of 260 > Full-text markdown of papers #10051–10100 of 12993 published in 2026-06, newest first. AI-readable mirrors of the full 12993-paper 2026-06 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-06/page/202 [← Previous bundle](https://aipod.dev/papers/archive/2026-06/page/201.md) [Next bundle →](https://aipod.dev/papers/archive/2026-06/page/203.md) ### TBD-VLA: Temporal Block Diffusion Vision Language Action Model Authors: Sung-Wook Lee, Xuhui Kang, Yen-Ling Kuo Published: 2026-06-05 Categories: cs.CV, cs.RO arXiv: https://arxiv.org/abs/2606.07895v1 Discrete Vision-Language-Action (VLA) models typically formulate action generation as next-token prediction over discretized action spaces, conditioning each token autoregressively on prior context. While effective, this paradigm incurs high inference latency and largely ignores the temporal structure inherent in action trajectories. Recent efforts introduce parallel decoding to improve efficiency, enabling faster inference, but lack explicit mechanisms for modeling token dependencies. We introduce TBD-VLA, a discrete token-based VLA framework that incorporates block diffusion to enable temporal action generation. We partition action sequences into temporal blocks and perform masked discrete diffusion within each block, while maintaining autoregressive generation across blocks. This design unifies temporal autoregression and parallel action decoding, achieving both strong temporal coherence and improved inference speed. In addition, the explicit temporal modeling enables asynchronous execution of action chunks (e.g., Real-Time Chunking) via temporal in-painting. TBD-VLA significantly outperforms prior VLA approaches in both simulation and real-world manipulation tasks, offering a scalable path toward fast, temporally aware, discrete VLA models. Project webpage: https://tbd-vla.github.io/ --- ### Beyond Individual Personas: Aligning Synthetic Dialogue to Population-Level Behavior Distributions Authors: Xinyi Liu, Rinat Khaziev, Hooshang Nayyeri, Emine Yilmaz, Charith Peris, Hari Thadakamalla Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07893v1 Synthetic dialogue corpora are increasingly used as proxies for target dialogue data, yet persona-grounded generators optimize individual conversations rather than corpus composition, yielding locally plausible dialogues with distorted population-level behavior mixes. We introduce GroupPersona, a framework that aligns synthetic dialogue corpora to the behavior distribution of a reference corpus. GroupPersona turns population statistics into generation controls: it separates each dialogue's core behavioral signature from predictable side effects, and uses the resulting behavioral groups to condition user agents on the interaction patterns that define the reference population. We evaluate GroupPersona on four corpora crossing two dialogue sources, assistant-style and Reddit-derived, with two construction variants: structure-preserving and variation-enhanced. GroupPersona lowers Jensen-Shannon divergence between synthetic and reference distributions over 12 behavior attributes from 0.234 to 0.177 relative to the strongest average baseline, a 24.4% reduction, and is best or tied-best on all four corpora while preserving structural alignment. It also achieves the closest calibration to reference-conversation quality scores, reducing mean absolute deviation from the reference-conversation profile to 0.63 versus 0.91 for the next-best baseline. --- ### C3VD-DEFCOL: A Deformable Colonoscopy Dataset with Time-Resolved 3D Ground Truth and Realistic Appearance Authors: Ethan Luk, Mayank V. Golhar, Anthony Song, Raúl Iranzo, Víctor M. Batlle, Lalithkumar Seenivasan, José M. M. Montiel, Nicholas J. Durr Published: 2026-06-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2606.07891v1 3D reconstruction could improve colonoscopy by estimating mucosal coverage and alerting clinicians to missed regions during screening. However, algorithm development is limited as no current datasets provide both a realistic in vivo appearance and dense, time-resolved 3D ground truth, especially under non-rigid deformation. We present C3VD-DEFCOL, a framework and dataset for evaluating deformable colonoscopy reconstruction with paired geometry and realistic texture. Starting from C3VD/C3VDv2 colon meshes and camera trajectories, we generate controlled deformations of the colon surface, including peristaltic waves and centerline motion, and render per-frame depth, surface normals, optical flow, camera poses, and time-stamped 3D meshes. We then use the rendered geometry, primarily depth, to condition an LTX-2.3-based sim-to-real translation model that produces RGB clips with in vivo-like mucosal color, texture, vasculature, and specular appearance while preserving the underlying 3D scene structure. The resulting dataset contains 110 videos from 11 unique colon mesh geometries, with varying camera trajectories, appearances, and parameterized deformation regimes, including three peristaltic severity levels that serve as controlled evaluation axes. We evaluate the generated videos using appearance realism, geometric consistency, and temporal consistency metrics, and use the paired ground truth to benchmark the downstream task of pose estimation in deformable 3D reconstruction. Our experiments show how pose estimation error increases with increasing deformation severity, providing a controlled stress test that is not possible with existing in vivo datasets. Overall, C3VD-DEFCOL is designed as a reproducible, quantitative evaluation platform for testing deformable 3D reconstruction algorithms, with the goal of reducing the domain gap between synthetic datasets and in vivo colonoscopy. --- ### Partially Performative Prediction Authors: Jaewook Lee, Tijana Zrnic Published: 2026-06-05 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2606.07890v1 Performative prediction studies feedback loops that arise when predictive models are deployed in consequential domains. In these settings, deploying a model can change the population whose patterns the model aims to predict, inducing a distribution shift that is endogenous to the learning system. This perspective departs from classical treatments of distribution shift, where shifts are typically modeled as exogenous changes in the data-generating process. Yet, in practice, distribution shift is rarely one or the other. Predictive models may influence future data through the decisions they support, while the world itself continues to drift for reasons beyond the learner's control. We study partially performative prediction, a framework that captures both endogenous and exogenous sources of distribution shift. The framework generalizes performative prediction by allowing the data distribution to evolve both in response to the deployed model and according to an external, time-varying process. We extend the central notions of performative stability and performative optimality to this setting by defining their online analogues that track the evolving partially performative environment. We analyze practical learning heuristics, including repeated retraining, and characterize when they successfully adapt to partially performative environments. --- ### Strained Coherence: A Pre-Failure Signal in Coding Agent Execution Trajectories Authors: Marut Pandya, Kasey Zhang, Baiqing Lyu Published: 2026-06-05 Categories: cs.LG, cs.AI, cs.CL arXiv: https://arxiv.org/abs/2606.07889v1 LLM-based coding agents sometimes acknowledge a problem in their own reasoning and then proceed anyway. We call this pattern strained coherence: a safety-relevant failure mode in which an agent has information that should change its behavior, states that information, and still acts against it. The pattern overlaps with verbalized reward hacking, where an agent names a tension between a task proxy and the underlying goal yet optimizes the proxy anyway. We give an operational definition, build a Claude Sonnet 4.6 judge that reads full trajectories and flags spans where the pattern occurs, and evaluate it on 44 Terminal-bench-2 trajectories using a Qwen3.5-35B-A3B backbone. Flagged trajectories fail 94% of the time versus 46% for unflagged trajectories (47-point gap, Fisher's exact p = 0.003; 46 points after excluding three prompt-embedded examples, p = 0.006). At matched selectivity, the detector reaches 94% precision versus 88% for a lexical discourse-marker baseline; the 10-trajectory intersection of the two methods has a 100% failure rate (Clopper-Pearson 95% CI (69%, 100%)). We replicate on Gemma4-31B with 43 trajectories: the overall signal is directionally consistent but not significant (20-point gap, p = 0.31), with attenuation driven largely by 13 trajectories with zero think content, where the detector has no substrate to analyze. In the high-verbosity Gemma tertile, the gap is +30 points; in the mid- and high-verbosity Qwen tertiles, it is +40 points each. The first flag appears at a median of 83-84% of elapsed trajectory time across both models, and the binary flag survives paraphrases that soften explicit conflict markers (8/8 trajectories). Unlike univariate predictors, the detector emits interpretable span-level output -- quoted acknowledgment, quoted action, and typed conflict -- showing what the agent saw and ignored. --- ### The Cross-Architecture Substrate: A Domain-Transcendent, Calibration-Surviving Geometric Invariant of Modern Vision Encoders Authors: Yousef Radwan Published: 2026-06-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2606.07882v2 Different vision neural networks -- trained to classify, contrast, reconstruct, or match images to text -- should have correspondingly different internal representations. We report that they do not. After training, the top sixteen principal directions of variation inside thirteen modern vision encoders converge to the same sixteen-dimensional geometric object. We call this the cross-architecture substrate and study it with PCA, centred kernel alignment (CKA), and Pang 2026 calibration. The substrate transports across four visual domains (natural photographs, medical CT, satellite, microscopy) at median Procrustes-CKA 0.679, and across eight domains (adding sketches, depth, thermal infrared, astronomy) at 0.604, every pair >0.40. It survives Pang calibration globally (7.4x disc-vs-MAE separation, n=13,394) and locally (4.82-5.30, p<10^{-44}). It is not pixel statistics (0.263), not Gabor features (0.31), not a random projection (0.041), and emerges in the first 10% of training while accuracy keeps climbing. We deliver four applications: a label-free transferability filter beating LogME (3x faster, +0.15 Kendall-tau); a four-way domain detector (99.6% accuracy); a frozen low-shot probe (16 dims beat 768-dim DINOv2 by 3.78pp at N=50 labels per class); and a teacher-free distillation auxiliary matching trained-teacher KD on 33 pairs (7.56pp peak gain at 10% label fraction). The substrate does not cross modalities, does not help cross-paradigm distillation, and does not predict transfer quality (rho=0.08 against transfer accuracy). --- ### Breaking the Bubble: Asynchronous Pipeline Parallel Training with Bounded Weight Inconsistency Authors: Itay Elam, Eliron Rahimi, Avi Mendelson, Chaim Baskin Published: 2026-06-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2606.07881v1 Pipeline parallelism is essential for training large neural networks, but existing schedules trade off throughput, memory, and optimization consistency. Synchronous pipelines preserve forward/backward weight consistency but suffer from bubbles; asynchronous pipelines remove bubbles but introduce weight-version mismatch, typically requiring weight stashing, prediction, or correction mechanisms. We introduce PACI (Pipeline Asynchronous training with Controlled Inconsistency), a bubble-free asynchronous pipeline method that bounds forward/backward version drift without weight stashing, prediction, additional parameter copies, or global synchronization. The key idea is to use local gradient accumulation as a version-control mechanism: by slowing parameter-version evolution relative to pipeline delay, PACI limits the number of optimizer updates crossed by any micro-batch while preserving steady-state utilization. In GPT-style language-model pretraining, PACI matches the stability and final perplexity of synchronous 1F1B-flush, retains the same peak memory footprint, achieves fully utilized pipeline throughput, and improves training time-to-accuracy by up to $1.69\times$ over the fastest flush baseline. These results show that forward/backward inconsistency need not be eliminated: when explicitly bounded, it can be safely traded for substantial efficiency gains. --- ### Still: Amortized KV Cache Compaction in a Single Forward Pass Authors: Charles O'Neill, Alex Sandomirsky, Harry Partridge, Mudith Jayasekara, Max Kirkby Published: 2026-06-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2606.07878v1 The KV cache is the memory bottleneck of long-horizon language model deployment. Practically, a deployable compactor must be lightweight enough to call during inference, expressive enough to preserve context under constraint, and reusable across a trajectory. Existing compaction methods satisfy only part of this requirement: selection methods are lightweight but subset-bound, while synthesis methods are expressive but rely on per-context optimization. Here we introduce Still, a small per-layer Perceiver trained once against a frozen base model that produces compact keys and values in a single forward pass. On Qwen and Gemma models, Still occupies the favorable side of the speed--quality frontier across compression ratios from $8\times$ to $200\times$ and context lengths from $8$k to $128$k. On the long-context RULER grid, Still exceeds the strongest baseline by 8--22 points. The same compact cache also supports free-form summarization, preserving most of the full-context gain on HELMET and winning a pairwise LongBench summarization comparison against KV-Distill. Because compaction is a forward pass, Still can be applied iteratively, entering a long-horizon regime unavailable to per-context methods. We show that amortization makes long-context cache compaction tractable, and synthesis makes its compact state useful at extreme compression. --- ### Whose Norms? Disentangling Cultural and Personal Alignment in Large Language Models Authors: Angana Borah, Isabelle Augenstein, Rada Mihalcea Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07877v1 Large language models are increasingly used for social decision-making situations that require balancing cultural norms with personal preferences. For example, a user preferring honesty might ask whether to correct a coworker publicly when local norms favor indirect feedback. Yet existing research studies cultural alignment and personalization largely separately. We introduce PACT, the Personal-Preference and Cultural-Norm Trade-off framework, which evaluates whether models choose to follow a cultural norm or allow personal preferences. We find that LLMs vary in how rigidly they enforce cultural norms, with behavior shifted more by country context (7.8%) than age (1%) and gender (0.7%) and shifting non-uniformly after instruction tuning. Furthermore, our five-country human study on PACT shows that culture-following in humans is mainly driven by scenario country, with the lowest agreement when participants judge their own cultural contexts, showing within-culture pluralism. Finally, human-LLM alignment experiments show that models can match majority choices, but fail to capture response distributions and uncertainty (with best correlations reaching only 0.24). Together, these findings motivate alignment evaluations that go beyond majority to capture cultural pluralism and disagreement in social judgment. --- ### Safety is Contextual, LLM-Judges Are Not: Navigating the Rigid Priors of Evaluators Authors: Anissa Alloula, Federico Licini, Ava Batchkala, Seraphina Goldfarb-Tarrant Published: 2026-06-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2606.07874v1 LLMs-as-judges are the only way to evaluate safety at scale. Despite their importance, LLM-judges themselves are rarely evaluated beyond human agreement in simple, static benchmarks. We therefore investigate two under-explored but crucial properties of LLMs-as-judges: their susceptibility to relying on in context-information, and their steerability to differing safety definitions, which may not align with their internal safety priors. We evaluate the safety judging abilities of many generalist LLMs and safety-specific judges, and investigate the impact of task demonstrations, novel in-context information, and changing safety definitions. We find that while LLM-judges can learn from new information, they are broadly unlikely to adjust their evaluations if the context or safety definition contradicts their prior. --- ### VisualFLIP: Do Predictions Depend on Task-Critical Visual Evidence in Multimodal Reasoning? Authors: Didi Zhu, Changrui Chen, Stefanos Zafeiriou, Jiankang Deng Published: 2026-06-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2606.07872v1 When a multimodal large language model answers a visual reasoning question correctly, is the prediction actually supported by the task-critical visual evidence? Correct answers can coexist with flawed reasoning, making accuracy alone an incomplete test of grounding. We introduce VisualFLIP, a paired benchmark with 1,374 images arranged as same-question perturbation pairs across cardinality, attribute, spatial, and logic tasks. Each pair keeps the question fixed but minimally changes the evidence so the gold answer deterministically flips. We evaluate 24 MLLMs with pair accuracy, which requires solving both sides of a pair, and Collapse Rate (CR), which measures how often a model that solves at least one side repeats the same non-empty answer for both images. Together, these metrics show that paired correctness and evidence dependence are related but distinct: capable models can still fail to update after task-critical visual changes, and collapse becomes more severe for some models when the edited image follows an earlier answer in a sequential setting. Further details are available on our project page: https://didizhu-judy.github.io/VisualFLIP/ --- ### The Cold-Start Safety Gap in LLM Agents Authors: Chung-En Sun, Linbo Liu, Tsui-Wei Weng Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07867v2 Are tool-calling LLM agents equally safe throughout a conversation? We discover they are not: agents are most vulnerable at the very start of a session and become substantially safer after a few regular agentic tasks -- a phenomenon we term the cold-start safety gap. To study this systematically, we introduce Safety Over Depth for Agents (SODA), a benchmark that controls how many regular agentic tasks the agent completes before encountering a safety threat, supporting up to 20 preceding tasks. Evaluating 7 models from 4 families, safety improves by 9--52% as the number of preceding regular agentic tasks increases from zero to twenty. Representation analysis confirms that model hidden states gradually shift toward a safety-aligned region as more preceding tasks are present. By systematically studying which part of the preceding conversation matters most, we find that the regular agentic tasks themselves are the primary driver of safety, while the agent's own prior responses have less effect on safety but are essential for preserving later utility. This conclusion is further supported by evaluation on open-source safety benchmarks (AgentHarm, Agent Safety Bench) and utility benchmarks (BFCL, API-Bank), confirming that warming up the agent with regular agentic tasks before deployment makes it safer and preserves full capability. Based on these findings, we recommend a simple deployment strategy: having the agent complete a few regular agentic tasks before possible exposure to safety-critical requests mitigates the cold-start safety gap. Our code is available at https://github.com/Trustworthy-ML-Lab/Agent-Cold-Start-Safety-Gap --- ### The Cold-Start Safety Gap in LLM Agents Authors: Chung-En Sun, Linbo Liu, Tsui-Wei Weng Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07867v1 Are tool-calling LLM agents equally safe throughout a conversation? We discover they are not: agents are most vulnerable at the very start of a session and become substantially safer after a few regular agentic tasks -- a phenomenon we term the cold-start safety gap. To study this systematically, we introduce Safety Over Depth for Agents (SODA), a benchmark that controls how many regular agentic tasks the agent completes before encountering a safety threat, supporting up to 20 preceding tasks. Evaluating 7 models from 4 families, safety improves by 9--52% as the number of preceding regular agentic tasks increases from zero to twenty. Representation analysis confirms that model hidden states gradually shift toward a safety-aligned region as more preceding tasks are present. By systematically studying which part of the preceding conversation matters most, we find that the regular agentic tasks themselves are the primary driver of safety, while the agent's own prior responses have less effect on safety but are essential for preserving later utility. This conclusion is further supported by evaluation on open-source safety benchmarks (AgentHarm, Agent Safety Bench) and utility benchmarks (BFCL, API-Bank), confirming that warming up the agent with regular agentic tasks before deployment makes it safer and preserves full capability. Based on these findings, we recommend a simple deployment strategy: having the agent complete a few regular agentic tasks before possible exposure to safety-critical requests mitigates the cold-start safety gap. Our code is available at https://github.com/Trustworthy-ML-Lab/Agent-Cold-Start-Safety-Gap --- ### Overcoming the Regulatory Bottleneck via Agent-to-Agent Protocols: A Nuclear Case Study Authors: Akshay J. Dave, David Grabaskas, Joseph A. Renevitz, Richard B. Vilim Published: 2026-06-05 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2606.07866v1 Regulatory review of advanced nuclear reactor designs routinely spans more than three years and consumes hundreds of millions of dollars in combined regulator and applicant labor. We present the Regulatory Context Protocol (RCP), an Agent-to-Agent communication standard that replaces the formal human-to-human pipeline between regulators and applicants with a structured, auditable agentic channel, while preserving human oversight at safety-significant decision points. The protocol is calibrated against an analysis of 1,236 documents from U.S. Nuclear Regulatory Commission advanced reactor dockets and demonstrated with a working multi-agent pilot. Against an 89M USD, 42-month Reconstructed Baseline, RCP cuts costs by 50-77 percent (21M-44M USD) and timelines by 65 percent (15 months). Without a shared protocol, Standalone Agents reach only 54M-74M USD and 21 months. The residual cost-and-time gap is structural, not algorithmic: it traces to the inter-organizational pipeline that only an agent-to-agent standard can compress. The same bottleneck - formal multi-party review under strict auditability requirements - characterizes pharmaceutical approvals, environmental permitting, financial supervision, and aviation certification. The US regulatory paperwork burden carries a 426.5 billion USD annual opportunity cost; replicated broadly, the projected 50-77 percent reduction implies savings on the order of 210-330 billion USD per year - approaching 1 percent of US GDP. --- ### Instrumented data for causal scientific machine learning Authors: Daniel N. Wilke Published: 2026-06-05 Categories: cs.LG, cs.AI, physics.comp-ph, stat.ML arXiv: https://arxiv.org/abs/2606.07865v1 Scientific machine learning is limited less by model size than by the data it is trained on. Observational data records what happened but not why; template synthetic data has a known generating process but only for the simulator's template, not the case a user faces. We argue a third option is now operationally feasible: instrumented data, in which every datum carries the mechanistic model that produced it, an explicit uncertainty over that model, and an executable family of counterfactuals. Verification-and-validation (V&V) instrumented image-to-simulation pipelines are one realisation: a sensor observation becomes a fully specified, solver-backed simulation with explicit, editable parameters and a propagated aleatoric/epistemic uncertainty. The substrate is case-specific, mechanistically supervised, and supports causal interventions through Pearl's do-operator. Near-term consequences for validation, auditing, and surrogate training span computational biology, climate, materials, fluid mechanics, and medical imaging; a longer-term, falsifiable implication concerns foundation models for scientific reasoning. --- ### The Last Visible Pixel: Probing Fine-Scale Perception in Vision-Language Models Authors: Lujun Li, Lama Sleem, Niccolo Gentile, Yangjie Xu, Yewei Song, Wenbo Wu, Radu State Published: 2026-06-05 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2606.07861v1 Recent vision-language models (VLMs) excel at multimodal understanding and reasoning, yet their fine-grained visual perception remains underexplored. A natural extension of ``How many r are there in Strawberry?'' asks: how small a visual pattern can a VLM reliably perceive? As such, we introduce FineSightBench, a new benchmark that systematically probes this limit by separating perception tasks (pixel-level recognition of letters, shapes, objects) from reasoning tasks (spatial reasoning, counting, ordering over small targets) across controlled scales of 4--48px. Through comprehensive experiments and detailed failure mode analysis on state-of-the-art models, we reveal a sharp dissociation: perception saturates around 12px, while reasoning remains limited even at larger scales, with persistent numeracy and sequence errors. These findings expose fundamental deficiencies in VLMs' fine-scale visual reasoning that demand more rigorous evaluation. --- ### Model Multiplicity for Adversarial Detection in Small Language Model Training on Edge Devices Authors: Stefan Behfar, Richard Mortier Published: 2026-06-05 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2606.07857v1 The rise of edge-based machine learning has enabled distributed adaptation of language models across mobile and IoT devices, offering privacy preservation and real-time responsiveness. However, distributed fine-tuning of language models on untrusted or heterogeneous edge nodes introduces new vulnerabilities. Compromised or unreliable devices can inject poisoned updates, leading to stealthy model manipulation or convergence degradation. Classical defenses such as robust aggregation or temporal anomaly detection operate on a single global model and are therefore limited in detecting coordinated or persistent poisoning. This work proposes a new system-level defense based on model multiplicity. Instead of maintaining one global model, the system rotates or concurrently trains multiple small language models (e.g., DistilGPT-2), each updated by independently sampled subsets of edge nodes. These models evolve under distinct training trajectories, creating multiple independent views of the same distributed population. Divergence between models quantified through gradient similarity, loss evolution, or parameter variance serves as a signal of anomalous or adversarial behavior. When one model deviates significantly from the ensemble mean, the system flags its contributing nodes for isolation or re-weighting. We implement this framework and evaluate it on edge-scale simulations of Small Language Model (SLM) training under varying heterogeneity and attack conditions. Results show that model multiplicity enables earlier and more reliable detection of poisoning compared to classical single-model defenses such as Flanders and Robust methods. Our findings demonstrate that diversity in model evolution can serve as a practical and effective defense mechanism for secure distributed learning on resource-constrained edge devices. --- ### Teacher-Free Self-Training Amplifies but Does Not Compound: A Pass@$K$ Crossover on a Free-Verifier Domain Authors: Igor Lima Strozzi Published: 2026-06-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2606.07856v1 When a language model trains on its own verified outputs, does it acquire capability beyond its base, or merely get better at expressing capability the base already had? We make the question decidable with a teacher-free "constellation" -- a generator, a learned critic, and a free exact verifier -- on a FlashFill-style "trapdoor" DSL, where verified (problem, solution) pairs are cheap to synthesize, hard to invert, and free to check exactly. Everything runs on one 4-bit Qwen3-4B on a single 24 GB GPU, with no model in the loop larger than the base. We report three findings. (i) Critic-guided selection beats verifier-filtered best-of-$k$ by $+9.1$ pp ($6/6$ seeds), with the entire gain localized to tasks where candidates disagree on held-out inputs. (ii) Per-round STaR self-training raises the ceiling but never accelerates -- the gain tracks remaining headroom and decelerates across $K=4$ independent training trajectories. (iii) The domain has no clean zero-capability frontier, so the usual "$0\% \to$ climb $=$ emergence" test is invalid here. A measured pass@$K$ crossover settles the diagnosis: the trained model wins at the operating budget (pass@$8$) but the base overtakes it at a large budget (pass@$64$) on every trajectory, so self-training concentrates probability mass rather than expanding reach. This is amplification, not compounding. ($K=4$ is indicative, not yet a robust across-trajectory CI.) --- ### Beyond English benchmarks: clinical llm evaluation in Brazilian Portuguese Authors: Giordano de Pinho Souza, Glaucia Melo, Josefino Cabral Melo Lima, Daniel Schneider Published: 2026-06-05 Categories: cs.CL, cs.AI arXiv: https://arxiv.org/abs/2606.07853v1 Large Language Models are transforming the support for clinical decision and their application in real scenarios. Yet, most benchmarks are conducted in English, and cross-lingual evaluation is needed to tackle the language gaps in global access. We introduce ClinicalBr, the first bilingual benchmark for clinical decision built from real Brazilian case reports. The corpus contains 2,892 cases drawn from 28 SciELO medical journals, spanning 18 specialties, and is structured as parallel Portuguese-English pairs. Each case supports four evaluation tasks: diagnosis retrieval, differential diagnosis, exam recommendation, and treatment planning. We evaluate four models: MedGemma-27B, Sabiá-4, DeepSeek-R1, and o3-mini, across both languages. The central finding is that the Portuguese-English performance gap is task-dependent, not general. In diagnosis retrieval, English yields a consistent advantage across all models, with +7.5-12.1 accuracy points. This advantage disappears in differential diagnosis, exam recommendation, and treatment planning, where confidence intervals cross zero for most models and Portuguese completeness scores are marginally higher. Brazilian-endemic conditions proved easier than the full corpus, not harder, indicating that tropical presentations are adequately represented in current pre-training. Exam recommendation was the hardest task across all models and both languages, with F1 scores below 0.10, well below the differential diagnosis ceiling of 0.20-0.27. --- ### Cost-Aware Speculative Execution for LLM-Agent Workflows: An Integrated Five-Dimension Method Authors: Faisal Fareed Published: 2026-06-05 Categories: cs.DC, cs.AI, cs.MA arXiv: https://arxiv.org/abs/2606.07846v1 LLM-agent workflows chain model calls and tool invocations, and spend most of their wall-clock time waiting on upstream operations before downstream ones can start. Speculative execution can reclaim that idle time by launching a downstream operation with a predicted upstream input, but here each speculation costs real money (per-token billing) and its success probability is hard to estimate and drifts over time. This paper presents a method organized around five design decisions: (D1) start a downstream operation before its upstream completes; (D2) price each speculation in real dollars at separate input and output rates; (D3) expose a single operator dial for latency versus cost; (D4) decide via an expected-value rule with a failure-weighted cost term and a preference-adjusted threshold; and (D5) estimate the success probability with a Bayesian Beta-Binomial posterior whose prior is keyed to a dependency-type taxonomy. Variants of these ideas appear in recent work; the combination, with every decision logged in dollars, is what is new. The rule fires only on edges passing an admissibility precondition (side-effect-free, idempotent, or stageable behind a commit barrier), since a wrong speculation is rolled back by re-execution, which refunds tokens but cannot un-send an irreversible side effect. We specify the runtime mechanics, a closed-form result that the rule self-limits as the upstream branching factor grows, a five-stage calibration pipeline (offline replay, shadow, canary, online calibration, drift-triggered kill-switch), and a workload-fit rubric over eight production archetypes. Contrast tables against the four closest published systems (DSP, Speculative Actions v2, Sherlock, B-PASTE) show differentiators on every dimension, and a synthetic validation suite confirms the predicted decision boundary, probability threshold, posterior recovery, and streaming-cancellation behavior. --- ### GRPO Does Not Close the Multi-Agent Coordination Gap Authors: Najmul Hasan, Prashanth BusiReddyGari Published: 2026-06-05 Categories: cs.MA, cs.LG arXiv: https://arxiv.org/abs/2606.07845v1 We measure how well current large language models coordinate as multiple agents sharing a common resource, using the dining philosophers problem as a clean test bed. Across 630 episodes spanning seven models and three philosopher counts, four frontier closed-source systems reach mean reward 0.45 to 0.87 and Mistral-Small 24B reaches 0.83 to 0.99, while Qwen3-14B reaches 0.13 to 0.35. We then ask whether group relative policy optimization (GRPO) on rollouts from the task itself can close the gap and find that it cannot: a Welch's t-test on per-episode reward at five philosophers gives p = 0.66 and a Hedges' g of -0.11, with no statistically significant change at ten or fifteen philosophers either. Two further observations qualify the result. The training reward of both 8B and 14B runs peaked at step nine and then declined, so the default saved checkpoint at step 15 is strictly worse than several earlier ones. The four-term reward we use admits a degenerate maximum at zero actions, which DeepSeek-R1-Distill-Qwen-7B and Mistral-Small 24B at five philosophers both inhabit, with mean reward 1.0 and 0.83 respectively at zero meals. The bottleneck for an open-weight 14B model on multi-agent coordination is not training compute but training methodology: reward shaping that does not collapse to a no-action maximum, checkpoint discipline that does not depend on the final step, and curriculum across problem scales. --- ### RACT: Retrieval Augmented Column-Table Learning and Prediction for Multi-Table Schema Matching Authors: Leonard Traeger, Enas Khwaileh, Andreas Behrend, George Karabatis Published: 2026-06-05 Categories: cs.DB, cs.IR, cs.LG arXiv: https://arxiv.org/abs/2606.07843v1 Schema matching, a critical task for integrating data from diverse sources, seeks to identify correspondences between columns across different schemas. In multi-table holistic schema matching, columns with similar semantic meaning may reside in tables with different contexts due to heterogeneous schema designs, where similarity-based techniques are inadequate. The focus of this paper is exploiting referential context into schema matching by introducing RACT learning and prediction, a self-supervised framework enabling the probabilistic retrieval of candidate tables for source columns to constrain relevant column candidates. Experiments demonstrate that this approach outperforms similarity-based baselines on matching multi-table schemas. In subsequent matching experiments, constraining the column search space via top-t tables improves both average matching precision and completeness by up to +70%. --- ### Large-scale empirical tuning and comparison of default optimizers for variational inference Authors: Trevor Campbell, Jonathan H. Huggins, Kyurae Kim, Charles C. Margossian Published: 2026-06-05 Categories: stat.CO, cs.LG, stat.ML arXiv: https://arxiv.org/abs/2606.07841v1 Black-box variational inference (BBVI) is a methodology for posterior approximation that relies on stochastic optimization. In practice, the stochastic optimizers underpinning BBVI generally require extensive problem-specific tuning, which undermines its promise as a truly "black box" inference algorithm. However, over the past decade, many new adaptive stochastic optimization algorithms have been developed that reduce or remove entirely the need for tuning. In this work, we investigate this new collection of adaptive methods in the context of BBVI, with the goal of establishing the current state of the art in tuning-free optimization-based inference. In particular, we present a large-scale empirical evaluation of 56 stochastic gradient-based optimization algorithms applied to 1092 Bayesian inference optimization problems, involving over 550,000 individual optimization runs and 15 core-years of compute. The optimization algorithms we evaluate are chosen to represent a wide spectrum of recent approaches and the benchmark problems are chosen to span a range of difficulty, with posterior target dimension 1-10^4, condition number 1-10^8, and a range of variational families. Our results show that no single method dominates, but running a selection of 5 algorithms suffices to reliably get close to the best-possible observed performance. We thus provide a strong baseline for applications where expert tuning is not possible and for comparison when developing new stochastic optimization algorithms. --- ### Does Persona Make LLMs K-pop Fans? A Pilot Study of LLM-Based Online Concert Audience Agents Authors: Kirak Kim, Hyojin Kim, Yejin Son, Sungyoung Kim, Kyung Myun Lee Published: 2026-06-05 Categories: cs.HC, cs.AI arXiv: https://arxiv.org/abs/2606.07837v1 A concert is a collective experience, but recorded performance videos are typically watched alone, stripping away the shared audience presence that makes concerts feel eventful. We investigate whether persona-based LLM audience agents can recreate aspects of this collective experience by generating real-time fan chat alongside a K-pop performance video. We present a multi-agent system in which ten LLM agents react through live-chat messages, comparing a persona-conditioned audience (each agent assigned a distinct fan identity, bias, and chat style) with a no-persona baseline. In a within-subjects pilot with K-pop fans (N=11), persona conditioning substantially improved model-level chat quality and perceived naturalness, but did not translate into differences in social connectedness, engagement, or affective response. Interviews suggest that online K-pop concert chat may operate as collective monologue rather than interpersonal dialogue, and that meaningful participation depends on shared identification with the specific artist and fandom. Persona conditioning can make LLM audiences appear more natural, but culturally meaningful collective experience may require deeper alignment between persona, crowd behavior, fandom identity, and user expectations. --- ### Agentic multi-fidelity learning of quasiparticle and excitonic properties Authors: Arnab Neogi, Aaron Forde, Christopher A. Lane, Sergei Tretiak, Jian-Xin Zhu Published: 2026-06-05 Categories: cond-mat.mtrl-sci, cond-mat.stat-mech, cs.AI, physics.comp-ph, quant-ph arXiv: https://arxiv.org/abs/2606.07836v1 Many-body GW-Bethe-Salpeter equation calculations are essential for accurate simulations of electronic structure and optical properties in modern low-dimensional nanomaterials. However, these methods are computationally demanding and can exhibit localized numerical instabilities or convergence failures that are difficult to detect within high-throughput workflows. We introduce an agent-guided multi-fidelity framework for correcting GW-Bethe-Salpeter excited-state landscapes in strained MoS2-WS2 bilayers. Across stacking registries, strain branches and reciprocal-space samplings, the workflow identifies spike-like excursions, near-zero-gap collapse and cross-fidelity inconsistencies associated with fragile long-wavelength dielectric screening. A structural agent evaluates calculations by assigning confidence weights and selectively using a small number of high-accuracy reference points. Machine learning models then transfer information across related systems and apply Gaussian process corrections to recover improved quasiparticle gaps and exciton binding energies, with calibrated uncertainty estimates. The approach corrects numerically induced artifacts without erasing physical strain dependence and substantially improves agreement with higher-fidelity references relative to a no-agent baseline. These results show that reliable surrogate learning for excited-state materials requires explicit diagnosis of numerical fragility, not direct interpolation of raw first-principles data points. The proposed framework is readily transferable to other optoelectronic nanomaterials characterized by strong quantum confinement, such as quantum dots, nanoribbons, layered two-dimensional semiconductors, and hybrid perovskite nanostructures. --- ### Mitigating the Contractivity Trap in Diffusion ODEs via Stein Stabilization Authors: Shigui Li, Delu Zeng Published: 2026-06-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2606.07835v1 A fundamental tension exists in the large-step inference of diffusion models via their deterministic probability flow ordinary differential equation (PF-ODE) trajectories, which we identify as the contractivity trap: efficient inference favors large step sizes, while aggressive steps and highly expressive denoisers can undermine contraction-based stability certificates for error suppression. To address this, we propose SteinDiff, a step-wise inference-time stabilization framework that employs Stein-derived corrections without requiring reference samples. Specifically, SteinDiff introduces a geometry-aware residual correction mechanism that regularizes large-step solver updates without retraining. To this end, we derive a closed-form Stein correction coefficient for step-wise solver adjustment, enabling reference-free adaptation to local data geometry. We further establish a score-controlled perturbation bound under distributional shifts and provide a complementary Stein perspective on EDM-style parameterizations. Extensive experiments demonstrate that SteinDiff mitigates severe artifacts and improves generative quality across large-step inference settings. --- ### Cherry-pick Override: LLM Judges Under-use the Non-Directional Verdicts Their Contract Authorizes Authors: Haoran Xu Published: 2026-06-05 Categories: cs.SE, cs.AI, cs.CL, cs.MA arXiv: https://arxiv.org/abs/2606.07834v2 Evidence-grounded fact verification increasingly uses LLM judges to turn evidence into terminal verdicts. Many task contracts deliberately include non-directional verdicts - Conflicting for materially mixed evidence, Not Enough Evidence for absent evidence - so that a system can decline to assert a direction. Returning Supports or Refutes on a claim whose authorized verdict is non-directional is directional overcommitment: a label error whose output is a direction the contract did not authorize. Its conflict instance - both strands present, one of them returned - is Cherry-pick Override (CPO); unlike the insufficiency instance it cannot be repaired by retrieving more, because nothing was missing. On AVeriTeC's gold-Conflicting subset (N_C=150) a four-option typed panel commits directionally on 18.7% of claims, and four frozen contemporary judges (two proprietary, two open-weight) on 21.3-48.0%, with 24.0-38.0% on VitaminC-Mixed, a constructed stress test on which the ranking reverses. These are conservative lower bounds: the shared prompt specifies Conflicting in detail, forbids collapsing to a direction, and demonstrates the label three times. The failure is not specific to conflict: overcommitment on gold-Insufficient cases runs at a comparable rate for every judge and exceeds CPO for three of four on VitaminC-Mixed. Asked in a fresh context, two judges identify the conflict on about two thirds of their own CPO claims, against under 0.12 where they committed correctly: the recognition is available and does not reach the terminal action. Typed vocabulary, panel voting, confidence thresholds, a structural validator and self-decomposition each leave a substantial residual, and the last buys part of its reduction by relabelling one-sided evidence as conflicting. Every rate is dataset-defined; we report no human-confirmed prevalence. --- ### Cherry-pick Override: Unsafe Directional Commitment in LLM Judges under Mixed Evidence Authors: Haoran Xu Published: 2026-06-05 Categories: cs.SE, cs.AI, cs.CL, cs.MA arXiv: https://arxiv.org/abs/2606.07834v1 LLM judges increasingly turn verdicts into system commitments. Under mixed evidence (claims with both supporting and refuting sources) this is unsafe: when the schema exposes CONFLICTING as the authorized non-directional verdict, returning SUPPORTS/REFUTES is an unauthorized directional commitment, a failure we name Cherry-pick Override (CCO). We define CCO under an explicit task contract and report it with a same-denominator diagnostic protocol paired with matched-coverage bootstrap and an apples-to-apples random-veto null. On AVeriTeC's Conflicting subset (N_C = 150), three-option judges return a directional verdict on more than 84% of mixed-evidence claims; under the typed schema, three-judge majority voting amplifies direction-on-conflict on AVeriTeC (0.887 vs. 0.840; 95% CI (+0.013, +0.080)) but does not replicate on VitaminC-Mixed. Walking an intervention ladder of common single-channel fixes (typed vocabulary, panel aggregation, confidence thresholding, validator-only filtering), each leaves a distinct residual failure: panel aggregation suppresses single-judge CONFLICTING dissent in 48% of CCO cases; the panel is well-calibrated for direction (ECE = 0.07 on pure-S/R) so confidence cannot operationally separate CCO from correct directional commits; validator-as-classifier nearly halves pure-evidence accuracy. A minimal two-channel reference probe reaches operating points neither single channel reaches; under the random-veto null its promotion to CONFLICTING is structurally targeted on AVeriTeC (empirical p < 1/2001) and weaker but in the same direction on VitaminC-Mixed, a selectivity result rather than a magnitude one. We argue for an external commitment-control layer that separates verdict generation from commitment authorization, using structural evidence and confidence as orthogonal channels and NO-COMMIT as a routed controller state. --- ### Beyond Pass/Fail: Using Process Mining to Understand How LLMs Resist (and Fail) Red Team Attacks Authors: Zvi Topol Published: 2026-06-05 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2606.07833v1 Standard AI red teaming evaluations reduce adversarial campaigns to a single binary outcome, attack success rate (ASR), not taking into account the sequential structure of how models resist or yield to attacks. We propose applying process mining, a discipline for discovering and analyzing process models from event logs, to red teaming traces. We conduct a controlled experiment pitting 60 HarmBench prompts against two LLMs, GPT-OSS 120B and Llama 3.3 70B, using 10 prompt mutation strategies over up to 110 attempts per prompt. From the resulting 8,575 scored events we extract Directly-Follows Graphs (DFGs) and state transition matrices that reveal structurally distinct defense profiles invisible to ASR alone: GPT-OSS exhibits a near-absorbing refusal state, while Llama presents multiple porous escape routes from refusal to getting successfully jailbroken. We further show that mutator effectiveness is asymmetric across models and that time-to-jailbreak distributions differ by an order of magnitude. --- ### Jas: AI-Paired Engineering as a Revival of N-Version Programming Authors: Jason Hickey Published: 2026-06-05 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2606.07828v1 I report a case study in AI-paired software engineering: five working ports of a vector illustration application across Rust, Swift, OCaml, Python, and browser-based platforms, built by a single developer in approximately 120 evening hours. The methodology pairs AI-assisted implementation with two safeguards -- a precise executable YAML specification serving as the single source of truth, and parallel implementations functioning as a built-in differential-testing layer. The five ports share a 23{,}000-line specification; per-port native code ranges from 0 to roughly 95{,}000 lines, reflecting the specification's escape hatch. I argue that AI-paired engineering, conditional on these two safeguards, makes feasible scope of work that conventionally requires multiple developer-years, and frame the methodology as a revival of N-version programming, a 1980s approach abandoned on cost grounds that AI changes. The paper reports concrete artifacts and honest limitations of the single-developer case study. --- ### The ACUTE Protocol: Operationalizing Language Model Activations for Better Calibration, Utility, and Trust Authors: Nishant Subramani, Palash Goyal, Yiwen Song, Mani Malek, Yuan Xue, Tomas Pfister, Hamid Palangi Published: 2026-06-05 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2606.07822v2 As language models improve and become increasingly deployed to solve a variety of tasks, trustworthiness becomes essential. Calibration is a good proxy for trust: well-calibrated confidence estimates help inform the risk versus reward tradeoff when trusting a specific model output. Unfortunately, even as models improve, they remain poorly calibrated, often biasing towards overconfidence. Additionally, calibration can be gamed: a policy that always predicts the base rate is perfectly calibrated, but completely uninformative. To resolve this, we develop a new metric, expected utility renormalized by the oracle (EURO), that balances calibration and informativeness. We also propose a general-purpose activation-based confidence, utility, and trust estimation protocol (ACUTE) to appropriately adjudicate uncertainty. The ACUTE protocol provides flexible, sample-efficient, and compute-efficient confidence estimators for 3 tasks including multiple choice question answering, tool-calling, and scientific document summarization across 6 models from 4 model families. ACUTE outperforms strong baselines on EURO, while maintaining low calibration error. Taken together, our work shows that equipping LLMs with the ACUTE protocol can improve calibration, utility, and trustworthiness in numerous settings. --- ### Joint Structural Pruning and Mixed-Precision Quantization for LLM Compression Authors: Hoang-Loc La, Truong-Thanh Le, Amir Taherkordi, Phuong Hoai Ha Published: 2026-06-05 Categories: cs.AI, cs.LG arXiv: https://arxiv.org/abs/2606.07819v1 Recently, the efficiency of Large Language Models (LLMs) deployment has become a critical concern in practical applications. While post-training quantization (PTQ) and structural pruning are established techniques for reducing memory footprint and inference latency, most existing PTQ approaches optimize quantization errors on a per-layer basis, overlooking how errors accumulate and propagate through the network, often resulting in suboptimal solutions. Traditional pipelines also tend to apply pruning and quantization in isolation or sequentially, further compounding sub-optimality. We introduce a novel end-to-end framework that addresses these limitations in two key ways. First, we propose a novel mixed-precision PTQ strategy that directly minimizes global error propagation across the entire model, rather than isolating layer-wise errors. Building on this, we develop a novel joint optimization approach that simultaneously learns structural pruning decisions and mixed-precision quantization policies within a unified search space. Extensive experiments show that, at ultra-low precisions (1-3 bits), our quantization method reduces WikiText perplexity by up to 21% compared to state-of-the-art (SoTA) weight-activation quantization baselines. Against leading weight-only quantization methods, it achieves up to 59% and 85% lower perplexity on WikiText and C4, respectively. Compared to the SoTA joint pruning-and-quantization techniques, our proposed method delivers superior perplexity and reasoning performance at ultra-low bits. --- ### Representational Similarity and Model Behavior in Multi-Agent Interaction Authors: Yujin Potter, Seun Eisape, Shiyang Lai, Alexander Huth, James Evans, Been Kim, Jacob Eisenstein, Dawn Song, Alane Suhr Published: 2026-06-05 Categories: cs.CL, cs.NE arXiv: https://arxiv.org/abs/2606.07818v1 Researchers have shown that neural similarity among humans predicts social closeness and cooperative success, whereas innovation often emerges from interactions among dissimilar individuals. We investigate whether these principles extend to artificial intelligence by examining interactions between large language models. In our experiments, 276 model pairs interact across eight games spanning both cooperation and novelty. We find that pairs with more similar representation spaces achieve significantly higher cooperation but exhibit reduced novelty and creativity. The effects of representational similarity on cooperation and novelty remain robust even after controlling for other factors such as performance disparity and model size. We also find that similarity in the early layers consistently shows the strongest association with cooperation and novelty, compared to the middle and later layers. This suggests that a central factor underlying these patterns could be the extent to which the two models share lexical and semantic grounding. Overall, representational similarity can be an important consideration in multi-agent system design. --- ### MinNav: Minimalist Navigation Using Optical Flow For Active Tiny Aerial Robots Authors: Aniket Patil, Mandeep Singh, Uday Girish Maradana, Nitin J. Sanket Published: 2026-06-05 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2606.07813v1 Navigation using a monocular camera is pivotal for autonomous operation on tiny aerial robots due to their perfect balance of versatility, cost and accuracy. In this paper, we introduce MinNav, a navigation stack based on optical flow and its uncertainty to fly through a scene with static and dynamic obstacles and unknown-shaped gaps without any prior knowledge of the scene components and/or their locations/ordering. We further improve success rate by using the activeness of the robot to move around in an exploratory way to find obstacles and navigate. We successfully evaluate and demonstrate the proposed approach in many real-world experiments in various environments with static and dynamic obstacles and unknown-shaped gaps with an overall success rate of 70%. To the best of our knowledge, this is the first solution to tackle all the aforementioned navigation cases without prior knowledge using a monocular camera. Our approach is on par in performance with depth based methods with factors of magnitude less computation required and can readily run onboard tiny aerial robots. The accompanying video, supplementary material, code and dataset can be found at https://pear.wpi.edu/research/minnav.html --- ### Scaling Participation in Modular AI Systems Authors: Shangbin Feng, Yike Wang, Weijia Shi, Luke Zettlemoyer, Yejin Choi, Yulia Tsvetkov Published: 2026-06-05 Categories: cs.AI, cs.CL arXiv: https://arxiv.org/abs/2606.07812v1 Humanity is a mosaic of multifaceted talents and needs, and any truly intelligent AI must reflect that richness. Yet the LLMs used by all are built by the few -- a centralized market of monolithic AI models structurally ill-suited to capture the diversity of human knowledge, reasoning, and values. Here we introduce scaling participation, a new paradigm in which modular AI systems are built from the bottom up through the contributions of diverse stakeholders. Participants contribute small models trained on their own interests and priorities; these models then collaborate in modular frameworks as compositional AI systems. Participatory AI systems outperform monolithic LLMs by up to 15.4% across 15 tasks, such as reasoning and factuality, surpassing models larger than all contributed components combined. Further experiments show that participatory AI systems benefit from contributor diversity, substantially improve on each contributor's original priorities, and exhibit emergent capabilities that allow them to solve over 15% of problems where all individual models fail. Scaling participation provides a technical foundation for transitioning from the monolithic status quo toward an open, bottom-up, and collaborative AI future. --- ### SLMJury: Can Small Language Models Judge as Well as Large Ones? Authors: Anish Laddha, Nitesh Pradhan, Gaurav Srivastava Published: 2026-06-05 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2606.07810v1 Large language models (LLMs) are widely used as judges for evaluating model outputs, but their high cost, latency, and opacity limit scalability. We introduce SLMJury, a framework for evaluating small language models (SLMs) as judges across two paradigms: closed-ended binary correctness and open-ended quality scoring. We benchmark 16 SLM judges (0.6B-14B parameters) from four model families across ten benchmarks: eight closed-ended tasks spanning mathematical, scientific, and general reasoning (N=64,824 judgments per configuration), plus SummEval and MT-Bench for summarization and conversational scoring. We formalize judging as a budget-conditioned function and study five dimensions. Four findings emerge. (1) The overthinking effect is domain-dependent: for most judges quick 10-token verdicts match or beat extended reasoning on mathematical judging (by 2-7% where they help), while reasoning wins on general tasks by up to 23%. (2) Domain generalization separates model families, with math-to-general accuracy gaps ranging from under 10% to nearly 40%. (3) Closed-ended and open-ended judging draw on different capabilities: the best binary judge (Phi-4) drops to rank 9 on MT-Bench, while reasoning-trained models invert this ordering. (4) Under the Reflect-Critique-Refine (RCR) debate protocol, multi-agent debate degrades accuracy across all tested configurations, whereas the top judges resist six adversarial personas with <=0.55% variance. Reliable automated evaluation does not require large proprietary models, yet no single SLM dominates. The leaderboard is available at https://anishh15.github.io/SLMJury/, and our framework code and pip package are publicly available at https://github.com/anishh15/SLMJury and https://pypi.org/project/slmjury/. --- ### Where Instruction Hierarchy Breaks: Diagnosing and Repairing Failures in Reasoning Language Models Authors: Sanjay Kariyappa, G. Edward Suh Published: 2026-06-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2606.07808v1 Reasoning language models deployed in agentic workflows must follow an instruction hierarchy: when instructions from different sources conflict, the model should obey the highest-privilege applicable instruction. Existing benchmarks largely measure this behavior end-to-end, asking whether the final response is compliant. However, a non-compliant response can arise from several distinct failures: the model may fail to identify the relevant instructions in context, fail to resolve conflicts among identified instructions, or correctly resolve the conflict in its reasoning while still producing a violating response. We introduce a white-box diagnostic framework that localizes instruction hierarchy failures into instruction identification, conflict resolution, and response realization, making failures more interpretable. We evaluate three reasoning models--Gemma-4-31B-IT, Qwen3.6-35B-A3B, and Claude Sonnet 4.6--on long-context adaptations of IHEval and IHChallenge, and find that the dominant failure mode varies across models, tasks, and context length. Building on the observation that models can often detect conflicts and output violations when explicitly prompted, we propose two training-free self-monitoring mechanisms: a parallel input monitor for low-latency conflict detection before generation, and a sequential output monitor for response-level review and repair. Across Gemma-4-31B-IT, Claude Sonnet 4.6, and GPT-5.3, the strongest monitor reduces rule-following non-compliance by 81-99%, with GPT-5.3 reductions of 86% under static attacks and 45% under adaptive attacks. --- ### Beyond Goodhart's Law: A Dynamic Benchmark for Evaluating Compliance in Multi-Agent Systems Authors: Yiyang Zhao, Zhuo Zhang, Qingxuan Le, Lizhen Qu, Zenglin Xu Published: 2026-06-05 Categories: cs.AI, cs.MA arXiv: https://arxiv.org/abs/2606.07805v1 The rapid evolution of Large Language Models (LLMs) from passive assistants to autonomous, execution-capable agents has introduced critical operational risks. Most current evaluation frameworks neglect procedural compliance, leading to ''Machiavellian'' behaviors where agents strategically violate safety rules to maximize rewards - a direct manifestation of Goodhart's Law. To address this blind spot, we introduce MAC-Bench, a dynamic, adversarial benchmark designed to evaluate the procedural alignment of multi-agent systems under realistic pressure. We propose the SERV(Seed - Evolve - Refine - Verify) pipeline, an ``Agent-as-a-Benchmark'' paradigm that transforms unstructured legal texts into executable, contamination-free scenarios. By synthesizing holographic sandbox environments and injecting calibrated social-engineering pressure vectors, MAC-Bench forces agents into Pareto-optimal trade-offs between task success and regulatory adherence. We introduced novel metrics: the Compliance-Weighted Success Rate (CSR) and the Machiavellian Gap (MG), and conducted a comprehensive evaluation of state-of-the-art frontier models to reveal the pervasive trade-offs between success and compliance. --- ### Memetic Capture: A Pluralistic Policy Framework for Governing AI-Driven Cultural Disempowerment Authors: Subramanyam Sahoo Published: 2026-06-05 Categories: cs.CY, cs.AI arXiv: https://arxiv.org/abs/2606.07802v1 Culture is the most insidious vector of gradual human disempowerment by AI: unlike economic or political displacement, cultural displacement attacks the very preferences and values through which humans recognise and resist disempowerment itself. We argue that existing AI governance frameworks suffer from a critical blind spot by treating cultural impact as secondary to economic and safety concerns. This paper develops \emph{memetic capture} as a unifying concept for AI-driven cultural disempowerment, and proposes the \textbf{Cultural Pluralistic Governance Framework (CPGF)}, a four-tier policy architecture combining quantitative cultural influence metrics, democratic value assemblies, pluralistic deployment standards, and transnational coordination mechanisms. We argue that pluralism is not merely an ethical requirement for such governance but a structural necessity: monocultural AI governance accelerates the very disempowerment it claims to prevent. We identify concrete policy levers, discuss implementation tensions, and outline a research agenda at the intersection of pluralistic alignment and cultural AI governance. --- ### Improving Multimodal Reasoning via Worst Dimension Optimization Authors: Haocheng Lv, Huaping Zhang, Qiuchi Li, Lei Li, Chunxiao Gao Published: 2026-06-05 Categories: cs.AI arXiv: https://arxiv.org/abs/2606.07801v1 Multimodal reasoning requires a path that retains integrity over a wide range of constraints, from visual grounding to logic consistency. However, the current Process Reward Models focus on heuristically defined rewards that equally weigh these factors, which may lead to the concealment of individual dimension failures by the dominating factors, without guaranteeing the validity of the reasoning process in general. --- ### Reconstructing and forecasting disease trajectories of patients with Alzheimer's disease using routine data in resource-constrained settings Authors: Ratnadeep Das, Atri Chatterjee, Sitikantha Roy Published: 2026-06-05 Categories: cs.AI, cs.LG, q-bio.NC arXiv: https://arxiv.org/abs/2606.07798v1 Alzheimer's disease is a progressive neurodegenerative disorder, and its progression varies substantially across patients. Existing work aims to forecast patients' future cognitive state, with minimal focus on reconstructing the state from past visits. Furthermore, in current research, quantifying predictive uncertainty remains underexplored and relies on costly modalities such as MRI, PET, and CSF, limiting their deployment in resource-limited settings. In this research, our primary objectives are: First, bidirectional prediction of cognitive scores from irregular visits to present the complete disease trajectory. Second, to enable interpolation and extrapolation capabilities to assist clinicians in informed prognostic decision making, and third, to provide a well-calibrated uncertainty estimate for all predictions, and finally, to achieve the objectives using the modalities available during routine visits. We propose a unified framework, GNOVA: A GRU-Neural ODE Variational Autoencoder. The architecture combines a Gated Recurrent Unit encoder and a Neural ODE decoder within a variational autoencoder framework. In our work, we forecast the CDR-SB and MMSE Scores. The GRU encoder allows for any number of inputs at any time point. The Neural-ODE decoder performs continuous estimation, allowing interpolation and extrapolation at any desired time point. The Variational autoencoder allows for uncertainty estimation in predictions. We worked with 1,727 patients from the ADNI dataset over 10 years; the model achieved mean absolute errors of 1.35 and 2.28 for CDR-SB and MMSE scores, respectively, without requiring any neuroimaging or biomarker data. Feature-ablation studies revealed that age, BMI, and APOE4 status were strong predictors. The proposed framework enables the reconstruction of incomplete patient histories and the anticipation of future cognitive states. --- ### MOLOT System Card: Malicious Operational Logic Observation Transformer Authors: Daniil Lopatkin, Maksim Mitrofanov, Stanislav Rakovsky, Aleksandr Khalikov Published: 2026-06-05 Categories: cs.CR, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2606.07792v1 MOLOT (Malicious Operational Logic Observation Transformer) is a static malicious-code detection system designed for SAST setup where package metadata, maintainer history, and dynamic execution traces may be unavailable or unreliable. The system represents source code as behavior sequences derived from static call graphs, includes an explanation stage that ranks suspicious behavior activities and maps them back to source-code locations. The approach is evaluated on Python and JavaScript packages from PyPI and npm, compared with opensource detection tools, and validated under product constraints including runtime, memory use, and false-positive rates observed in a real moderation workflow. We also release Open Malicious-Code Bench, a public benchmark for reproducible evaluation of malicious-package detection methods. The results show that static behavior-sequence modeling can provide accurate, explainable, and deployable malicious-code detection for modern DevSecOps workflows. --- ### Frequency-Scale Saliency for Spectral Descriptor Analysis in 3D Shape Retrieval Authors: Jianru Shen Published: 2026-06-05 Categories: cs.GR, cs.CV, cs.IR arXiv: https://arxiv.org/abs/2606.07791v1 Classical spectral descriptors such as the Heat Kernel Signature and Wave Kernel Signature are widely used for non-rigid 3D shape retrieval, yet their failure modes remain poorly understood. We present a frequency-scale saliency framework that audits these descriptors by quantifying the retrieval-level contribution of each descriptor scale interval through ablation. We introduce class spectral fingerprints to characterize category-level scale dependence, and show that descriptor similarity between class pairs is substantially correlated with retrieval failure, with a Spearman correlation of 0.479. Experiments on SHREC'11 demonstrate that short scales dominate retrieval performance while long scales are harmful, that HKS and WKS exhibit distinct scale dependence patterns, and that saliency-weighted retrieval improves mAP on hard categories by 0.156, with cross-fold and random-weight controls confirming that the gain is stable and not due to arbitrary reweighting. --- ### Byzantine Cheap Talk: Adversarial Resilience and Topology Effects in LLM Coordination Games Authors: Aya El Mir, Martin Takáč, Salem Lahlou Published: 2026-06-05 Categories: cs.LG arXiv: https://arxiv.org/abs/2606.07790v1 Multi-agent LLM systems increasingly rely on communication protocols for coordination, yet their robustness under adversarial and structural constraints remains poorly understood. Building on prior work showing that cheap-talk channels enable cooperation in LLM coordination games, we investigate two vulnerability classes in a 4-player Stag Hunt across six model families and 720 trials. First, when Byzantine agents signal cooperation but defect, non-Byzantine agents detect the betrayal within one round yet fail to adapt collectively: a substantial fraction continue cooperating despite repeated exploitation, unable to recover coordination due to the game's unanimity payoff structure. Second, explicitly restricting communication topology collapses cooperation, while applying identical restrictions silently preserves near-perfect cooperation. This establishes that coordination failure stems from agents' meta-reasoning about hidden information, not information loss itself. We identify two stable behavioral archetypes that replicate across all model cohorts: Defection-Prone models that switch permanently after betrayal, and Cooperation-Persistent models that continue cooperating at significant individual cost. These findings reveal concrete security vulnerabilities: communication channels can be exploited as adversarial injection vectors, and disclosing network topology to agents can degrade coordination even without any adversary present. --- ### A Framework for Evaluating and Benchmarking Concept Drift Detection Methods Authors: Vitor Cerqueira, Heitor Murilo Gomes, Marco Heyden, Bernhard Pfahringer, Albert Bifet Published: 2026-06-05 Categories: cs.LG, stat.ML arXiv: https://arxiv.org/abs/2606.07789v1 Data stream mining is fundamentally challenged by concept drift, where distributional changes can degrade model performance. Despite the proliferation of drift detection methods, progress in the field is hindered by inconsistent evaluation practices: studies rely on oversimplified synthetic data generators, adopt incompatible metrics, and lack transparency in hyperparameter selection, making fair comparisons difficult. We address this gap with a novel benchmarking framework comprising three contributions: (1) a drift simulation method that injects controlled distributional changes into real-world datasets via Monte Carlo trials, enabling supervised evaluation while preserving real-world data complexity; (2) an evaluation protocol for drift detection with timing-aware criteria, including the derivation of new metrics (e.g., F1 detection score, normalized detection time) that are comparable across streams; and (3) we advocate for a leave-one-dataset-out hyperparameter optimization protocol for drift detection methods that promotes configuration robustness across heterogeneous stream dynamics. We benchmark 14 widely used drift detection methods on 7 realworld datasets across 4 drift types (class prior, label swap, feature permutation, feature filtering), each under both abrupt and gradual transitions. Our experimental results provide insights into the strengths and weaknesses of current drift detection approaches while establishing baseline performance metrics for future research in this area. All code and experiments are publicly available. --- ### Evaluating RAG Reliability under Clean, Misleading, and Mixed Retrieval Authors: Sevgi Yigit-Sert Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07783v1 Retrieval-Augmented Generation (RAG) is widely used to improve the factual reliability of large language models (LLMs) by grounding answers in retrieved evidence. In misinformation-rich environments, however, retrieved content may include plausible but incorrect information, raising concerns about the reliability of RAG-based information access systems. In this work, we propose an evaluation protocol to systematically test how the RAG system handles conflicts between parametric knowledge and evidence retrieved from context with varying amounts of misleading information. We target correct answers to factoid questions that the model responds to correctly, even when there is no retrieval, and use this to test the system with clean, poisoned, and mixed evidence. The proposed analytical framework combines parametric override and confidence metrics to assess when and how misleading information affects the generation process of LLMs. This study aims to provide insights into the robustness of RAG systems in information disorder scenarios. --- ### Non-Archimedean Polydisc Spaces and Applications to Optimisation Authors: Paul Lezeau, Yiannis Fam, Anthea Monod, Yue Ren Published: 2026-06-05 Categories: math.OC, cs.LG, math.MG arXiv: https://arxiv.org/abs/2606.07782v1 We propose a new framework for optimisation over non-Archimedean spaces inspired by Berkovich geometry. Specifically, we introduce polydisc spaces, which consists of products of closed balls over a non-Archimedean field. These spaces retain the rigid hierarchical structure of the non-Archimedean field whilst acquiring many desirable geometric features absent from it. We show that metric trees embed naturally into these spaces, demonstrating their capacity to represent hierarchical data. We study their metric geometry, establishing properties such as geodesic uniqueness, confirming their comaptibility with classical optimisation techniques. We further propose a class of real-valued functions given by linear combinations of absolute values of polynomials. These functions admit a piecewise polynomial description along geodesics and satisfy a universal approximation property. We formulate a theory of optimisation on polydisc spaces: we prove existence of minimisers and explore algorithms for finding them. We provide an accompanying open-source Julia library implementing the core objects and optimisation procedures introduced. --- ### Land cover and flood type govern the detection limits of satellite-based flood mapping across diverse global flood events Authors: Venkatesh Kolluru, Rajat Shinde, Abdelhak Marouane, Caden Helbling, Deepak Shah, Othneil Drew, Iksha Gurung, Manil Maskey, Rahul Ramachandran Published: 2026-06-05 Categories: cs.AI, cs.CV, cs.LG arXiv: https://arxiv.org/abs/2606.07780v1 Floods are among the most destructive natural hazards, and their increasing frequency under climate change makes satellite-based inundation mapping essential for disaster response. Geospatial foundation models pretrained on satellite archives offer geographic transferability, but their operational reliability across diverse, unseen events remains uncharacterized. Here we deploy Prithvi-EO-2.0 across 19 out-of-distribution flood events (2017-2025) spanning six continents, eight climate zones, and six flood mechanisms, validating against two independent reference products. Detection accuracy depended jointly on land cover and flood type, with cropland yielding the highest agreement (IoU=52%) and riverine events the strongest detection (F1=0.69), while tree cover and built-up areas showed near-zero detection (IoU=4%) regardless of flood mechanism. Dual-reference validation revealed that apparent model error partly reflects definitional inconsistency between reference products rather than detection failure. Iterative pipeline testing identified 23 failure modes, with pipeline engineering dominating initial error over model capacity. These findings establish environment-dependent detection boundaries for operational satellite flood mapping. --- ### Unlocking Latent Value: Taxonomy-Guided Recovery of High-Performing Data from Low-Tier Web Corpora Authors: Neeraj Varshney, Sanket Lokegaonkar, Nasser Zalmout, Qingyu Yin, Priyanka Nigam, Bing Yin Published: 2026-06-05 Categories: cs.CL arXiv: https://arxiv.org/abs/2606.07778v1 Dominant web data curation pipelines for pretraining collapse document quality into a single composite score, systematically missing high-value content along dimensions the scorer underweights. We present a taxonomy-driven framework that recovers this value by filtering along semantically meaningful dimensions that composite scores fail to capture. First, building on the ESSENTIAL-WEB taxonomy, we introduce two novel dimensions: timeliness and cultural specificity, both of which show low pairwise NMI with existing ones. We annotate 14M documents using Qwen2.5 32B and distill into a lightweight 0.5B model. To enable rapid corpus-wide annotation, we additionally train a 73M multi-task MLP on E5 embeddings, achieving 50x inference throughput. Second, to navigate the combinatorial explosion of filter configurations, we introduce a compute-efficient two-pass framework: Pass 1 identifies the strongest dimension signals at small scale; Pass 2 constructs and evaluates conjunctive and disjunctive compound filters from the top performers - identifying high-performing configurations at a fraction of full scaling-law cost. Applying the selected filters to deprioritized web data, taxonomy-filtered subsets outperform their unfiltered baselines and even surpass the highest-quality tier. On mid-tier data, our best filter improves over its unfiltered baseline by 12.1% on reasoning, 9.5% on coding, and 2.0% on knowledge benchmarks, exceeding unfiltered top-tier data by 6.7% on reasoning and 13.7% on coding. Furthermore, filtered data from two tiers below the typical production threshold improves by 22.3% on reasoning and 19.5% on coding over its unfiltered baseline, surpassing top-tier data on coding benchmarks. These results establish that vast latent value remains locked in deprioritized web data, and that multi-dimensional taxonomy filtering is a principled, compute-efficient key to unlocking it. --- ### DALE-CT: Depth-Aware 2D Slice Encoders Learn an Anatomical World Model of Chest CT Authors: Evan W. Damron, Mahmut S. Gokmen, Mitchell A. Klusty, Caroline N. Leach, Emily B. Collier, V. K. Cody Bumgardner Published: 2026-06-05 Categories: cs.CV arXiv: https://arxiv.org/abs/2606.07775v2 Chest CT is among the highest-volume imaging exams in medicine, yet expert voxel-level annotations are scarce and costly, motivating encoders that learn directly from unlabeled scans. We present DALE-CT, a family of 2D slice-based Vision Transformers trained from scratch on chest CT with the heuristics-free LeJEPA objective. We introduce depth-aware slab sampling, which draws self-supervised views from across a physical $z$-axis slab rather than a single slice, implicitly tasking the 2D encoder with representing how anatomy changes between neighboring slices. The frozen representations trace each scan as a smooth anatomical trajectory, recover cranio-caudal slice ordering without labels, and distinguish slices by the anatomy they contain rather than by position alone. This anatomical world model emerges without any 3D or positional supervision, and an otherwise-identical encoder trained on slices in isolation never develops it. Building on this backbone, we introduce dense auxiliary supervision into the pretraining objective, using anatomical and abnormality masks to supervise patch and slice tokens alongside the self-supervised loss, and we compare the resulting variants against a DINOv2 baseline continually pretrained on CT-RATE. Among nine public and in-house models evaluated under the same protocol, DALE-CT-2S is the strongest 2D model in-domain, reaching 0.825 Macro AUROC on CT-RATE, within 0.024 of COLIPRI-CRM and without any text supervision. We subsequently scale the supervision-free configuration to a $\sim$287k-scan multi-source pool, to our knowledge the largest reported chest-CT pretraining corpus. The resulting DALE-CT-0-L posts the best 2D external-transfer point estimates, and we release it as our recommended backbone with the full model family, training code, and evaluation pipeline. ---