# Research Papers — 2026-07 · bundle page 89 of 222 > Full-text markdown of papers #4401–4450 of 11075 published in 2026-07, newest first. AI-readable mirrors of the full 11075-paper 2026-07 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-07/page/89 [← Previous bundle](https://aipod.dev/papers/archive/2026-07/page/88.md) [Next bundle →](https://aipod.dev/papers/archive/2026-07/page/90.md) ### Isaac Sim-to-Real: Reinforcement Learning based Locomotion for Quadrupeds Authors: Jordan Dowdy, Jean Chagas Vaz Published: 2026-07-20 Categories: cs.RO, cs.LG arXiv: https://arxiv.org/abs/2607.18135v1 Learning-based approaches to locomotion have risen in popularity in recent years, showing the capability for complex legged locomotion and whole-body control. Reinforcement learning (RL), the primary learning-based approach for locomotion, often utilizes a high-performance simulation tool, providing a controlled and efficient training and development environment. However, policies that perform well in simulation frequently encounter unexpected challenges when deployed on a physical system, known as the sim-to-real gap. This work presents a robust RL locomotion framework capable of whole-body control. The proposed RL framework utilizes Nvidia's new set of simulation tools, Isaac Sim, and its companion RL framework, Isaac Lab, for training, achieving a zero-shot sim-to-real policy. The performance of our policy is validated on physical hardware using the Unitree Go1, with experimental results showing similar velocity tracking performance to the quadruped's integrated controller, with a greater ability to recover from large disturbances, and achieve linear velocities of 2.0 m/s and angular velocities of 1.8 rad/s. --- ### Manifold-Constrained Hyper-Connections for Parameter-Efficient Finetuning Authors: Valentijn Oldenburg, Floris de Kam, Bente Zuijdam, Lieve Eberson, Nicky van Zutphen, Stef de Wildt, Ivo Verhoeven Published: 2026-07-20 Categories: cs.LG arXiv: https://arxiv.org/abs/2607.18130v1 Most parameter-efficient finetuning (PEFT) methods adapt weights or activations, thus leaving one of the key Transformer components unchanged: residual connections. This paper investigates Manifold-Constrained Hyper-Connections (mHC), a generalisation of residual connections, as a novel PEFT approach, wrapping frozen OLMo-2 backbones with learned residual routing modules. We find that mHC can finetune frozen Transformers, but that its role differs fundamentally from the original pre-training setting: in finetuning, fixing the residual mixing matrix to identity often improves performance. As a standalone PEFT method, mHC does not consistently outperform LoRA. However, at matched trainable parameter budgets, mHC+LoRA combinations improve language-modelling loss and show task-dependent benchmark gains at both 1B and 7B scale. Overall, our results identify residual routing as a distinct and promising novel PEFT axis. --- ### ClouDens: Operational Context-Aware Anomaly Detection for Large-scale Cloud System Monitoring Authors: Thu T. H. Doan, Mohammad Saiful Islam, Andriy Miranskyy, Ngoc-Thanh Nguyen, Rogardt Heldal, Patrizio Pelliccione Published: 2026-07-20 Categories: cs.NI, cs.LG arXiv: https://arxiv.org/abs/2607.18127v1 With the rapid growth of cloud computing infrastructures in scale and complexity, network monitoring for Large-scale Cloud Systems (LCSs) has become increasingly challenging, requiring automated and reliable anomaly detection to maintain service availability. Modern LCSs continuously generate telemetry logs from distributed cloud services, producing high-dimensional multivariate time series that capture system operations. Detecting anomalies in this context is difficult due to extreme dimensionality, complex dependencies among distributed components, and severe sparsity from intermittently active services. Taking these challenges into account, we first conduct an empirical study on telemetry logs from the IBM Cloud Console platform, and then propose ClouDens, an anomaly detection framework tailored to LCS monitoring that leverages operational-context attributes encoded in the telemetry log schema to improve detection accuracy and early identification of anomalies. ClouDens partitions high-dimensional telemetry logs into domain-guided subsets, constructs a context-aware graph modeling operational service dependencies, and employs Spatio-Temporal Graph Neural Networks for forecasting-based anomaly detection. We evaluate ClouDens on the recently released IBM Cloud Telemetry Dataset and provide practical insights into designing reliable anomaly detection solutions for LCS monitoring. Results show ClouDens achieves higher NAB scores in count-based telemetry features, indicating more accurate, earlier anomaly detection with broader coverage than a GRU-based model. Our study further reveals that telemetry feature subsets, operational-context modeling, scoring strategies, and sparsity imputation all substantially influence detection performance, offering practical guidance for designing and fairly benchmarking anomaly detection approaches for LCS monitoring. --- ### COVAriance-Induced Fairness Gap Penalty for Subgroup-Fair Clustering Authors: Kyungseon Lee, Hankyo Jeong, Kunwoong Kim, Kwanho Lee, Yongdai Kim Published: 2026-07-20 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2607.18119v1 Fair clustering aims to make cluster assignments independent of sensitive attributes, but this goal becomes challenging when multiple sensitive attributes jointly define many subgroups. In such settings, directly extending existing fair clustering algorithms is computationally expensive or numerically unstable, especially when the number of subgroups grows exponentially and some subgroups contain only a few instances. To address these challenges, we define a subgroup-fairness gap for clustering and derive a covariance-based surrogate that exactly matches this gap. We then introduce a continuous relaxation of the surrogate, enabling efficient gradient-based optimization and yielding our proposed algorithm, COVA-FC. We also show that subgroup fairness alone does not imply marginal fairness, and extend our framework to capture a subgroup-marginal-fairness gap. Experiments on benchmark datasets show that COVA-FC achieves competitive cost-fairness trade-offs and improves computational efficiency over existing baselines in both subgroup and higher-order marginal settings. --- ### SGA: Plug&Play Geometric Verification for Educational Video Synthesis Authors: Jhon Lopez, Carlos Hinojosa, Bernard Ghanem Published: 2026-07-20 Categories: cs.AI, cs.CV, cs.GR, cs.MA, cs.MM arXiv: https://arxiv.org/abs/2607.18116v2 Recent work leverages Large Language Models (LLMs) to generate executable code for pedagogical animations using libraries such as Manim. However, ensuring spatial correctness and visual legibility remains challenging, as existing frameworks emphasize pedagogical content while overlooking geometric occlusions. We propose the Symbolic Geometric Agent (SGA), a plug-and-play module for code-centric animation pipelines that intercepts LLM-generated code, performs partial execution to extract symbolic scene graphs, and applies targeted refinement when spatial conflicts are detected. We further introduce the Manim Visual Quality Score (MVQS), a deterministic rendering-free proxy for spatial integrity. Experiments on the MMMC-Code benchmark across four LLM backbones and two agentic pipelines show that SGA achieves a peak MVQS of 73.11 (Code2Video + GPT-5.1), corresponding to a 16.1% relative improvement over the raw baseline, and improves MVQS in 7 of 8 backbone x pipeline configurations. Additionally, we conduct a human evaluation showing that these improvements translate into human preference, with raters preferring SGA over the raw baseline in 84.4% of comparisons and over a VLM-based critic in 65.0%. --- ### SGA: Plug&Play Geometric Verification for Educational Video Synthesis Authors: Lopez Jhon, Hinojosa Carlos, Ghanem Bernard Published: 2026-07-20 Categories: cs.AI, cs.CV, cs.GR, cs.MA, cs.MM arXiv: https://arxiv.org/abs/2607.18116v1 Recent work leverages Large Language Models (LLMs) to generate executable code for pedagogical animations using libraries such as Manim. However, ensuring spatial correctness and visual legibility remains challenging, as existing frameworks emphasize pedagogical content while overlooking geometric occlusions. We propose the Symbolic Geometric Agent (SGA), a plug-and-play module for code-centric animation pipelines that intercepts LLM-generated code, performs partial execution to extract symbolic scene graphs, and applies targeted refinement when spatial conflicts are detected. We further introduce the Manim Visual Quality Score (MVQS), a deterministic rendering-free proxy for spatial integrity. Experiments on the MMMC-Code benchmark across four LLM backbones and two agentic pipelines show that SGA achieves a peak MVQS of 73.11 (Code2Video + GPT-5.1), corresponding to a 16.1% relative improvement over the raw baseline, and improves MVQS in 7 of 8 backbone x pipeline configurations. --- ### How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs? Authors: Prakhar Gupta, Terry Jingchen Zhang, Florent Draye, Bernhard Schölkopf, Zhijing Jin Published: 2026-07-20 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.18114v2 Modern LLMs are alarmingly susceptible to surprisingly simple immaterial changes of input prompts: a casual hint, an incorrectly labeled few-shot example, or a fake prior assistant turn often flips an originally correct answer. We study where this susceptibility, spanning sycophancy and related cue-induced biases, lives inside the model. Across five model families and seven bias types, we extract a per-bias direction from hidden states and triangulate it through three measures: probing, leave-one-dataset-out (LODO) transfer, and causal intervention. The susceptibility is largely shaped by alignment tuning rather than pretraining: pretrained base models generally cave much less to these biases, and their activations carry much weaker cue-specific signal beyond question content. Within aligned models, each bias has a coherent linear direction that we can both decode and steer along, recovering the unbiased answer across every family we test. The biases do not collapse into a single shared representation, however: cross-bias overlap is model-specific, and even behaviorally similar biases occupy different directions. The same intervention also provides a proof-of-concept debiasing tool, recovering a meaningful share of bias-induced errors while preserving most correct answers across all instruct families. Cue-induced bias is therefore best understood not as a single flaw in LLMs but as a family of causally effective linear directions that are largely shaped by alignment tuning. --- ### How Does Alignment Tuning Shape Representations of Sycophancy and Related Cue-Induced Biases in LLMs? Authors: Prakhar Gupta, Terry Jingchen Zhang, Florent Draye, Bernhard Schölkopf, Zhijing Jin Published: 2026-07-20 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.18114v1 Modern LLMs are alarmingly susceptible to surprisingly simple immaterial changes of input prompts: a casual hint, an incorrectly labeled few-shot example, or a fake prior assistant turn often flips an originally correct answer. We study where this susceptibility, spanning sycophancy and related cue-induced biases, lives inside the model. Across five model families and seven BCT bias types, we extract a per-bias direction from hidden states and triangulate it through three measures: probing, leave-one-dataset-out transfer, and causal intervention. The susceptibility is largely installed by alignment tuning rather than pretraining: pretrained base models barely cave to these biases, and their activations carry no cue-specific signal beyond question content. Within aligned models, each bias becomes a single coherent direction that we can both decode and steer along, recovering the unbiased answer across every family we test. The biases stay representationally distinct, however: cross-bias entanglement is model-specific rather than a property of the bias category, and even behaviorally similar biases occupy different directions. The same intervention also serves as a modest debiasing tool, recovering a meaningful share of bias-induced errors while preserving most correct answers across all instruct families. Cue-induced bias is therefore best understood not as a single flaw in LLMs but as a family of distinct, causally active directions that alignment tuning installs. --- ### Occlusion-Aware Panoptic Segmentation with Joint Position Embedding and Occlusion-Level Attention Authors: Wenbo Wei, Jun Wang, Shan Raza, Abhir Bhalerao Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18112v2 Panoptic segmentation in complex scenes remains challenging because of occlusions, yet modern approaches often neglect occlusion modelling. In this paper, we propose Position Embedding Modulation with Occlusion Level Attention (PEMOLA), a novel occlusion-aware module that can be seamlessly integrated into transformer-based panoptic segmentation. To obtain occlusion cues, we train an occlusion classifier on the COCO-OLAC dataset. The classifier derives the occlusion-level attention, which serves as spatial guidance, while the occlusion labels are encoded into a learnable embedding to produce channel-wise weights. Through joint modulation, PEMOLA elegantly introduces the occlusion priors into the position embedding, thereby improving the occlusion modelling. We further annotate the Cityscapes dataset with occlusion levels, termed Cityscapes Occlusion Labels for All Computer Vision Tasks (Cityscapes-OLAC), following the same labelling protocol as COCO-OLAC, to evaluate the cross-dataset generalisation ability of PEMOLA. Extensive experiments on COCO-OLAC and Cityscapes-OLAC demonstrate that PEMOLA consistently improves panoptic segmentation quality while introducing minimal computational overhead. These results highlight the importance of occlusion modelling, where incorporating occlusion-level attention helps deliver robust panoptic segmentation under occlusion. Code and dataset are available at https://github.com/wenbo-wei/PEMOLA. --- ### LLM-as-a-Coach: Experiential Learning for Non-Verifiable Tasks Authors: Tianzhu Ye, Li Dong, Guanheng Chen, He Zhu, Xun Wu, Shaohan Huang, Furu Wei Published: 2026-07-20 Categories: cs.LG, cs.CL arXiv: https://arxiv.org/abs/2607.18110v1 Reinforcement learning (RL) on open-ended tasks compresses an LLM's rubric-based evaluation into a scalar reward, discarding rich textual feedback and conflating responses with distinct quality profiles. We propose Experiential Learning (EL), which repurposes the feedback model from an LLM-as-a-Judge into an LLM-as-a-Coach. The coach distills its assessment of each on-policy response into transferable experiential knowledge, which conditions a teacher model and is internalized by the policy through on-policy context distillation. Compared with scalar rewards, this higher-bandwidth feedback channel provides dense supervision and preserves fine-grained preferences among high-quality responses. Across two policy families, with feedback from the policy itself or a proprietary model, EL consistently outperforms rubric-based RL on held-out and unseen open-ended tasks. Notably, EL generalizes better beyond the training distribution, and mitigates reward hacking. These findings establish experiential knowledge as a richer and more generalizable learning signal for post-training on non-verifiable tasks. --- ### SpEmoC: A Balanced Speaker-Segment Multimodal Emotion Benchmark Authors: Sania Bano, Shahzad Ahmad, Santosh Kumar Vipparthi, Sukalpa Chanda, Subrahmanyam Murala Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18109v1 Understanding human emotions in spoken conversations is a key challenge in affective computing, with applications in empathetic AI, human computer interaction, and mental health monitoring. However, existing datasets vary in scale, emotion distribution, modality alignment, and data partitioning strategies, which can influence reliable cross-dataset generalization and minority-emotion modeling. We introduce SpEmoC a Speaking segment Emotion for Conversations comprising 306,544 raw clips from 3,100 English language movies and TV series. From these, 30,000 high quality, class balanced clips are curated, featuring synchronized visual, audio, and textual modalities annotated for seven emotions through a hybrid pipeline that integrates pretrained models with human validation. SpEmoC uses strict movie- and series-level splits to prevent content overlap between split sets, allowing more reliable evaluation of model generalization. The dataset also maintains a near-balanced distribution across seven emotions, including minority classes such as Fear and Disgust, which supports more balanced learning across categories. Extensive experiments, including in-domain benchmarking, cross-dataset transfer, low-data training, class-imbalance analysis, and modality transfer show that balanced data and careful splitting lead to more stable performance across emotions when models are evaluated on other datasets. These results highlight the importance of dataset design for robust and transferable multimodal emotion recognition. --- ### FinSAgent: Corpus-Aligned Multi-Agent RAG Framework for Evidence-Grounded SEC Filing Question Answering Authors: Jijun Chi, Zhenghan Tai, Hanwei Wu, Tung Sum Thomas Kwok, Hailin He, Zixing Liao, Bohuai Xiao, Chaolong Jiang, Jianliang Lei, Jerry Huang, Peng Lu, Muzhi Li et al. Published: 2026-07-20 Categories: cs.IR, cs.CL, cs.MA arXiv: https://arxiv.org/abs/2607.18102v2 Financial question answering over U.S. Securities and Exchange Commission (SEC) filings requires retrieving and synthesizing heterogeneous evidence dispersed across long, standardized, and highly redundant disclosures. Existing retrieval-augmented and multi-agent systems typically derive retrieval queries directly from the user's question and rank candidates by semantic similarity. Together, these choices create prior-corpus misalignment: a mismatch between model priors and the target filings' structure, terminology, and evidence standards. As a result, query generation misses corpus-specific evidence, while semantic reranking favors topically similar but evidentially invalid false-positive chunks. We propose FinSAgent, an evidence-grounded multi-agent framework that reframes SEC filing QA as corpus-aligned retrieval planning and corrects both ends with a single principle: inject corpus-side conditioning wherever model priors would otherwise dominate. FinSAgent combines (1) role-specialized agents anchored to the mandated 10-K item structure, (2) database-aware query decomposition that conditions each agent's sub-queries on a lightweight, summary-level view of the local corpus, and (3) multi-path retrieval with a learned feature-gated reranker that separates evidential validity from semantic similarity. Across five offline financial QA benchmarks, FinSAgent improves retrieval coverage and answer correctness over strong single-agent and multi-agent baselines; in a three-arm randomized online experiment with 1,000 anonymous user ratings, it also receives higher scores than baselines. --- ### Empowering On-Device Model Adaptation with an Edge AI Inference Accelerator Authors: Mateusz Piechocki, Alessandro Capotondi, Marek Kraft Published: 2026-07-20 Categories: cs.LG, cs.AR, cs.CV arXiv: https://arxiv.org/abs/2607.18101v1 On-device model adaptation is essential to enable lifelong personalization on resource-constrained hardware, but compute, power, and memory limitations of such devices make end-to-end backpropagation impractical for modern deep neural networks. This work proposes a heterogeneous adaptation pipeline that repurposes a commercial edge AI inference accelerator, Hailo-8L, for frozen-backbone feature extraction during on-device training. The computational graph is partitioned so that the pre-trained backbone is quantized to INT8 and run on the accelerator, while only a lightweight FP32 classification head is fine-tuned on the host CPU, enabling frequent, energy-efficient in-field updates with most weights remaining fixed. Across multiple architectures and datasets, this pipeline achieves up to 15.4x faster wall-clock training time compared to a Raspberry Pi 5 CPU baseline, offers competitive throughput in favorable settings, and consistently reduces energy per sample. Post-training quantization restoration is shown to be crucial for preserving the quality of accelerator-generated features and mitigating accuracy loss in quantization-sensitive architectures. Overall, the results demonstrate a practical approach to efficient on-device adaptation using inference-oriented edge accelerators. The implementation is available at https://github.com/MatPiech/accelerator-training. --- ### Can We Break LLMs Out of Self-Loops? Fine-Grained Reasoning Control with Activation Steering Authors: Sheldon Yu, Tong Yu, Xunyi Jiang, Rohan Surana, Gagan Mundada, Sungchul Kim, Lina Yao, Julian McAuley, Junda Wu Published: 2026-07-20 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.18100v1 Extended reasoning has become standard for frontier Large Language Models (LLMs), yet the trajectories these models produce remain largely uncontrollable. Existing methods for shaping how a model reasons are prompt based approaches and operate at the input level, offering no fine-grained control over the reasoning process itself. Related work analyzes and discovers latent transition dynamics in the reasoning traces from Large Language Models. Building on this, we statistically characterize these states, and show that failure trajectories get stuck in self-loops, exhausting the token budget without progress toward the final answer. To intervene on these failures, We propose SOPHIA: Steering Of reasoning Processes via Hidden-state Intervention and Activations. We treat each reasoning trace as a sequence of latent states rather than an unstructured texts, and investigate whether inference time interventions can provide fine-grained control over the self-looping reasoning process. We classify every prefix to a latent state, record step level transitions, and use them to construct a bank of steering vectors indexed by state pairs. At inference time, a controller infers the current state and, given a target state, retrieves the corresponding vector and can also detect self-loops online from the transition structure to prevent the model from sinking into a reasoning black hole. Through extensive experiments, our method reliably intervenes on self-loop failures, with steering vectors that generalize to different state pairs. End task accuracy and token efficiency indicate that fine-grained controllability results in better reasoning quality. --- ### VDAR-Router: Adaptive LLMs Routing via Verbalized Query Difficulty Analysis Retrieval Authors: Yu-Chien Tang, Jun-Chen Hung, Wen-Chih Peng, An-Zi Yen Published: 2026-07-20 Categories: cs.CL arXiv: https://arxiv.org/abs/2607.18098v2 Large language models are increasingly used in practical systems, making efficient model selection important for reducing deployment cost. LLM routing has emerged as a practical solution for allocating each input query to an appropriate model under a desired cost-performance trade-off. Existing routing methods often estimate model suitability from the surface semantics or embedding similarity of the input query. However, such methods may ignore the underlying difficulty of a query, leading to suboptimal routing decisions. To address the challenge, we propose VDAR-Router, a difficulty-aware retrieval-based routing framework. For each input query, VDAR-Router first generates an explicit difficulty analysis. It then retrieves historical examples with similar difficulty profiles. Based on the retrieved records, it estimates candidate model suitability and selects the model using a reward function that considers both performance and cost. Experiments on three datasets show that VDAR-Router consistently achieves better cost-performance trade-offs than existing baselines. These results demonstrate the effectiveness of difficulty-aware retrieval for training-free LLM routing. Case studies further show that explicit query analysis helps retrieve more relevant examples and supports more reliable routing decisions. --- ### VDAR-Router: Adaptive LLMs Routing via Verbalized Query Difficulty Analysis Retrieval Authors: Yu-Chien Tang, Jun-Chen Hung, Wen-Chih Peng, An-Zi Yen Published: 2026-07-20 Categories: cs.CL arXiv: https://arxiv.org/abs/2607.18098v1 Large language models are increasingly used in practical systems, making efficient model selection important for reducing deployment cost. LLM routing has emerged as a practical solution for allocating each input query to an appropriate model under a desired cost-performance trade-off. Existing routing methods often estimate model suitability from the surface semantics or embedding similarity of the input query. However, such methods may ignore the underlying difficulty of a query, leading to suboptimal routing decisions. To address the challenge, we propose VDAR-Router, a difficulty-aware retrieval-based routing framework. For each input query, VDAR-Router first generates an explicit difficulty analysis. It then retrieves historical examples with similar difficulty profiles. Based on the retrieved records, it estimates candidate model suitability and selects the model using a reward function that considers both performance and cost. Experiments on three datasets show that VDAR-Router consistently achieves better cost-performance trade-offs than existing baselines. These results demonstrate the effectiveness of difficulty-aware retrieval for training-free LLM routing. Case studies further show that explicit query analysis helps retrieve more relevant examples and supports more reliable routing decisions. --- ### SciForma: Structure-Faithful Generation of Scientific Diagrams Authors: Yuxuan Luo, Peng Zhang, Xinjie Zhang, Xun Guo, Zhouhui Lian, Yan Lu Published: 2026-07-20 Categories: cs.CV, cs.GR, cs.LG arXiv: https://arxiv.org/abs/2607.18091v1 Structural fidelity is essential to scientific methodology diagrams. To communicate research logic, these diagrams must faithfully render components, directional relations, and textual annotations. Since a single error, such as a reversed arrow or an unreadable equation, can invalidate the entire figure, structural fidelity is inherently conjunctive: correctness on one axis cannot compensate for failure on another. Current open-source models fail to satisfy this criterion. Supervised fine-tuning (SFT) learns plausible layouts but cannot reliably ensure structural correctness, while scalar reward-based post-training obscures which structural dimension has failed. To address this, we introduce SciForma, a framework for the structure faithful generation of scientific methodology diagrams. Specifically, SciForma decomposes diagram quality into three structural axes: Component, Arrow, and Text, guided by a structural inventory. Built on this foundation, we curate SciFormaData-700K for structured training and SciFormaBench-2K for logic-verified evaluation. To close the gap left by SFT, we develop Multi-Dimensional Conjunctive Preference Optimization (M-DPO), which enforces simultaneous correctness across all axes and adaptively routes gradients to the most deficient dimension in post-training. The same structural inventory also enables iterative editing at inference time to correct residual errors. This combination allows SciForma-9B to exceed all open-source baselines and GPT-Image-1.5 on both SciFormaBench-2K and AIBench, bringing open scientific diagram generation close to proprietary-level structural fidelity. Our code and data will be available at: https://github.com/microsoft/SciForma. --- ### The Label Complexity of Class-Conditional Coverage under Distribution Shift Authors: Weijia Han, Lisha Qu Published: 2026-07-20 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2607.18088v2 Conformal prediction certifies that a classifier's prediction sets cover the truth, and that certificate is marginal. Many recognition benchmarks build distribution shift into evaluation, placing disjoint conditions in the training and test splits. Under that shift the certificate stays reassuring while per class coverage fails silently: on a real cross subject skeleton benchmark marginal coverage holds near ninety percent while the worst class is covered about seventy percent and ten of sixty classes fall below eighty percent. This class specific undercoverage stays hidden behind a single reassuring marginal number. Once the shift acts jointly on covariates and labels, the target class conditional score law is unidentified, so no label free method is at once per class valid and efficient uniformly over target laws consistent with the observed source joint distribution and target covariate marginal. The per class labels needed to recover every class threshold to a given tolerance grow as the inverse square of that tolerance and the logarithm of the class count, with matching bounds for classwise threshold procedures. Pseudo labels do not shortcut it: the best prediction powered estimator gains at most a small constant factor where coverage collapses. Across three real shifts and an image corruption benchmark, source label calibration recovers much of the gap while marginal coverage holds, and stops once it breaks. --- ### Judge-dependent safety gains and model-specific helpfulness costs of evidence-sufficiency prompting in clinical LLMs Authors: Koyar Afrasyab Published: 2026-07-20 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.18086v1 Background: LLM judges increasingly score whether clinical language models give overconfident answers under incomplete evidence, yet whether a measured "safety gain" reflects real behavior change or the judge's calibration is unresolved. Using a structured evidence-sufficiency prompt as a test case, we asked whether it reduces unsafe overconfident answers, how far that effect depends on the scoring judge, and what it costs in helpfulness. Methods: In a retrospective public-data benchmark (Real-POCQi, HealthBench, MedRBench), four models (GPT-5.5, Claude Opus 4.8, Gemini 3.5 Flash, Grok 4.3) answered a fully paired common panel (1,200 cells) with a standard prompt and the wrapper. The pre-specified endpoint was the paired reduction in unsafe overconfidence scored by the primary judge (GPT-5.4-nano); secondary analyses added a different-family judge (Claude Sonnet 5), a correctness judge, matched scaffold controls, and a blinded three-clinician review. Results: Unsafe overconfidence fell from 49.3% to 24.7%, a paired reduction of 24.7 points (95% CI 21.8-27.7; p<0.001), robust in direction across models and paraphrases. Magnitude was judge-dependent: Sonnet agreed on direction but nearly halved the effect (+13.1 points), with one-directional disagreement. Blinded clinicians characterized the primary judge as a high-sensitivity (1.00), low-specificity (0.55) screen, not a calibrated rate. The gain carried a model-specific helpfulness cost (correct diagnosis 80.3% to 50.3%): near-free for GPT-5.5, near-total for Gemini (-58 points). Matched scaffold controls showed genuine behavior change, not judge circularity. Conclusions: LLM-judged clinical safety effects should be reported as directional and relative, anchored to human review and evaluated jointly with helpfulness, not as calibrated absolute rates. This does not establish clinical deployment readiness. --- ### WorldCupArena: Fine-Grained Evaluation of Language Models and Deep-Research Agents on Football Forecasting Authors: Zhaokai Wang, Tianlin Gui, Jiayuan Rao, Shangzhe Di, Yihong Tang, Dingli Liang Published: 2026-07-20 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2607.18084v2 Predicting a football match before kickoff requires more than knowing past results: a model must use changing information and make a clear prediction before the answer is available. We present WorldCupArena, a dynamic benchmark for language models and deep-research agents. The 2026 FIFA World Cup is its first evaluation, and the same process can be reused for future leagues and cups. Before each match, a model either receives a common evidence package or searches for information itself. It predicts the result and score, likely players and events, match statistics, and the outcome of the competition. After the match, these predictions are compared with the recorded result. We report result accuracy, exact-score accuracy, and a scoreline score that gives some credit when a predicted score is close but not exact, together with scores for the other prediction tasks. Across systems, similar result accuracy can mask larger differences in detailed predictions. Four systems predicted champion Spain, and two of them also recovered the exact final pairing. Compared with betting-market and human-fan baselines, the best system shows only small gains in result and exact-score accuracy, but a clearer gain in Scoreline. New schedules can be added as they begin, allowing the benchmark to evaluate future models without using outcomes that are already known. Code, predictions and evaluation scripts will be publicly released. --- ### WorldCupArena: Fine-Grained Evaluation of Language Models and Deep-Research Agents on Football Forecasting Authors: Zhaokai Wang, Tianlin Gui, Jiayuan Rao, Shangzhe Di, Yihong Tang, Dingli Liang Published: 2026-07-20 Categories: cs.AI, cs.CL, cs.LG arXiv: https://arxiv.org/abs/2607.18084v1 Predicting a football match before kickoff requires more than knowing past results: a model must use changing information and make a clear prediction before the answer is available. We present WorldCupArena, a dynamic benchmark for language models and deep-research agents. The 2026 FIFA World Cup is its first evaluation, and the same process can be reused for future leagues and cups. Before each match, a model either receives a common evidence package or searches for information itself. It predicts the result and score, likely players and events, match statistics, and the outcome of the competition. After the match, these predictions are compared with the recorded result. We report result accuracy, exact-score accuracy, and a scoreline score that gives some credit when a predicted score is close but not exact, together with scores for the other prediction tasks. Across 104 matches and 13 systems, models with similar result accuracy differ more clearly on detailed predictions. Compared with betting-market and human-fan baselines, the best system shows only small gains in result and exact-score accuracy, but a clearer gain in Scoreline. New schedules can be added as they begin, allowing the benchmark to evaluate future models without using outcomes that are already known. Code, prompts, predictions, and evaluation scripts are open sourced at https://github.com/wzk1015/WorldCupArena. --- ### CriPO: Enhancing Rubric-based RL via Self-Distillation Authors: Mingxuan Xia, Yuhang Yang, Chao Ye, Shuai Zhu, Shenzhi Yang, Guangcheng Zhu, Yuhang Zhang, Cheng Peng, Haobo Wang, Chenglong Wang Published: 2026-07-20 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.18082v4 Rubric-based Reinforcement Learning (RL) has recently shown promise in improving Large Language Models (LLMs) on open-ended tasks. A widely recognized limitation of rubric-based RL is limited exploration: criteria that no rollout manages to satisfy (Unexplored Criteria) receive no optimization signal. Recent methods address this by incorporating rubric information as external guidance during rollout generation, yet they introduce a train-inference mismatch: the policy is optimized on rollouts produced under external guidance while this guidance is absent at inference time, causing error accumulation through autoregressive decoding. Moreover, these exploration-focused approaches overlook a fundamentally different failure mode that we term Suppressed Criteria---criteria that are satisfied by some rollouts yet whose learning signals might be lost during optimization because scalar reward aggregation assigns them non-positive aggregate advantages. Our analysis reveals that suppressed criteria constitute a persistent and non-negligible failure mode---over 25% of samples exhibit this issue throughout training. To simultaneously address both unexplored and suppressed criteria without introducing training-inference mismatch, we propose Criterion-Distilled Policy Optimization (CriPO), which enhances rubric-based RL via on-policy self-distillation. For unexplored criteria, CriPO constructs a behavior-injection teacher and computes a filtered forward-KL loss to inject missing behaviors into the policy. For suppressed criteria, CriPO uses a counterfactual teacher to locate criterion-relevant tokens in negative-advantage rollouts, and corrects their advantages in GRPO to preserve useful patterns. Experiments on medicine and science benchmarks demonstrate that CriPO outperforms existing rubric-based RL methods, e.g., achieving an average gain of 3.3 points over GRPO on Qwen3-4B. --- ### CriPO: Enhancing Rubric-based RL via Self-Distillation Authors: Mingxuan Xia, Yuhang Yang, Chao Ye, Shuai Zhu, Shenzhi Yang, Guangcheng Zhu, Yuhang Zhang, Cheng Peng, Haobo Wang, Siqing Wang Published: 2026-07-20 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.18082v3 Rubric-based RL has recently shown promise in improving LLMs on open-ended tasks. A widely recognized limitation of rubric-based RL is limited exploration: criteria that no rollout manages to satisfy (Unexplored Criteria, UC) receive no optimization signal. Recent methods address this by incorporating rubric information as external guidance during rollout, yet they introduce a train-inference mismatch: the policy is optimized on rollouts produced under external guidance while this guidance is absent at inference time, causing error accumulation through autoregressive decoding. Moreover, these exploration-focused approaches overlook a fundamentally different failure mode that we term Suppressed Criteria (SC) -- criteria that are satisfied by some rollouts yet whose learning signals are lost during optimization because scalar reward aggregation assigns them non-positive aggregate advantages. Our analysis reveals that SC are remarkably prevalent: over 57% of samples exhibit this failure mode throughout training, with an average of 1.8 SC per sample. To simultaneously address both UC and SC without introducing training-inference mismatch, we propose Criterion-Distilled Policy Optimization (CriPO), which enhances rubric-based RL via on-policy self-distillation. For UC, CriPO constructs a criterion-injection self-teacher and computes a localized forward-KL loss to inject missing behaviors into the policy. For SC, CriPO employs a counterfactual self-teacher to locate criterion-relevant tokens in negative-advantage rollouts and flips their token-level advantages to positive values, preserving useful patterns that would otherwise be suppressed. Experiments on medicine and science benchmarks demonstrate that CriPO consistently outperforms rubric-based RL, achieving stronger final performance with approximately $2\times$ fewer optimization steps. --- ### SelectInfer: Selective Neuron Loading and Computation for On-Device LLMs Authors: Huzaifa Shaaban Kabakibo, Eric Schniedermeyer, Artem Burchanow, Lin Wang Published: 2026-07-20 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.18081v1 Large Language Models (LLMs) have demonstrated remarkable capabilities across a range of Natural Language Processing (NLP) tasks, but their high computational and memory demands pose significant challenges for deployment on resource-constrained edge devices. Existing approaches to model compression and optimization often rely on coarse-grained pruning or quantization, which can compromise accuracy or require re-training and fine-tuning. In this work, we introduce SelectInfer, a neuron-level optimization framework that enables efficient LLM inference on edge devices through selective neuron loading and computation. By profiling and identifying both task-specific and general-purpose neurons using an offline LLM profiler, SelectInfer implements two key optimizations: selective loading, which reduces memory footprint by selectively loading a subset of neurons that were identified to be most important during the offline stage, and selective computation, which dynamically computes only the most relevant neurons at runtime. Evaluation across multiple datasets shows that SelectInfer achieves significant reductions in memory footprint and computation while preserving task performance, making it a practical step towards enabling LLM deployment on edge devices --- ### Sparse Evidence Can Suffice: Agentic Evidence Seeking for Multimodal Video Misinformation Detection Authors: Haochen Zhao, Yongxiu Xu, Xinkui Lin, Dong Xie, Jiarui Lu, Yuqi Qian, Yubin Wang, Hongbo Xu, Gaopeng Gou Published: 2026-07-20 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2607.18080v2 Multimodal video misinformation detection is commonly formulated as a holistic video-understanding task, where the entire video and its associated content are processed and judged in a single pass. However, real-world misinformation often exhibits a sparse and compositional evidence structure: a reliable decision may depend on only a few coupled clues, while most video content contributes limited additional information. Exhaustive multimodal reasoning may therefore introduce substantial redundancy and obscure decisive evidence. This motivates decoupling evidence acquisition from verification: first identifying sparse, decision-relevant clues and then judging veracity based on the acquired evidence. Accordingly, we propose SIEVE, a framework for Sparse Interactive Evidence Verification via Extraction in multimodal video misinformation detection. An evidence-seeking agent actively explores the available multimodal evidence and constructs a compact evidence package, which is then used by a verifier to determine veracity. The agent is trained with supervised evidence-seeking trajectories and an evidence-aware reinforcement learning objective that promotes informative evidence acquisition while discouraging unnecessary or invalid interactions. Experiments on multiple video misinformation benchmarks show that SIEVE consistently outperforms the evaluated baselines and supports reliable verification using compact evidence packages. Moreover, the resulting acquisition process provides an explicit and inspectable evidence trail, improving the transparency and groundedness of multimodal misinformation detection. --- ### VGOcc: Learning Visual-Geometric Gaussians for Vision-Centric 3D Driving Occupancy Prediction Authors: Junhong Lin, Xianda Guo, Kangli Wang, Yuqi Ye, Xiaoyu Liang, Yanlun Peng, Wei Gao Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18078v1 Vision-only occupancy prediction requires recovering a semantic 3D occupancy field from calibrated surround-view images, where each view provides observations with ambiguous depth along camera rays. Existing methods have progressed from dense structured representations to sparse Gaussian primitives, improving the efficiency of 3D scene representation. However, Gaussian learning still relies primarily on image domain features, which provide limited explicit geometric information for volumetric reasoning. Our key observation is that effective Gaussian occupancy modeling requires not only sparse primitives, but also richer geometric and semantic learning cues. In this paper, we propose VGOcc, which learns visual and geometric cues from foundation models for Gaussian modeling. VGOcc incorporates these cues into primitive initialization and refinement, yielding a representation termed Visual-Geometric Gaussians tailored to semantic occupancy prediction. Specifically, we propose Visual-Geometric Gaussian Birth to form spatially balanced Gaussian centers from ray depth hypotheses, while visual semantic features initialize primitive attributes. Next, we design Pose-Aware Feature Learning to combine foundation tokens with camera embeddings and calibrated ray information. Features from neighboring views are then aggregated at projected 3D locations for each Gaussian refinement stage. Finally, Gaussian decoder refines birth Gaussians with pose-aware features and renders them into semantic occupancy. Experiments on nuScenes demonstrate that VGOcc achieves state-of-the-art performance in vision-only 3D occupancy prediction. Codes will be available at https://github.com/JHLin42in/VGOcc. --- ### Generalised Bellman recurrence and three dualities in sequential decision-making Authors: Fernando E. Rosas, David Hyland, Daniel Polani Published: 2026-07-20 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.18077v1 What gives the Bellman equation its form? We show that the recursive properties of optimal value functions follow from three conditions: that the dynamics decomposes through sufficient statistics, that the return decomposes recursively, and that the aggregation of uncertainty is compatible with both. When all three conditions hold on a common state, the Bellman equation arises from their mutual consistency; when one fails, tractability can often be recovered by augmenting the state or by deforming return or dynamics. The same conditions are shown to give rise to three dualities: one between probability and return, one between return and aggregation, and one between aggregation and probability. Our framework reveals these dualities as arising from a single construction, unifying methods developed separately across reinforcement learning, control, and decision theory. --- ### Modeling turn-taking with distant viewing: investigating silence thresholds in human and AI-generated discourse Authors: Taylor Arnold, Nicolas Ballier, Artem Saloev Published: 2026-07-20 Categories: cs.CL arXiv: https://arxiv.org/abs/2607.18076v1 This study investigates silence gaps in two kinds of audiovisual material. We analysed thirty US situational comedies and fifty-one synthetic podcasts generated with Google NotebookLM. Gaps were compared across speaker gender, assigned from a fundamental-frequency threshold estimated in Praat, and across production settings. --- ### Sobek: Streaming Equivariant Tensor Product Convolutions Authors: Vladimir Chorošajev, Cédric Bény Published: 2026-07-20 Categories: cs.LG arXiv: https://arxiv.org/abs/2607.18074v1 Equivariant graph neural networks repeatedly apply edge-conditioned tensor-product convolutions over graph edges. Conventional implementations materialize edge-specific weights, messages, and adjoints, causing tensor-product workspace and memory traffic to grow rapidly with graph size and operator width. This limits feasible workloads and can prevent larger problems from fully utilizing the GPU. We show that these edge-sized intermediates are artifacts of the execution schedule, not requirements of the equivariant operator. By reassociating radial projection, spherical-harmonic coupling, and graph aggregation, edge-local products can be consumed directly into bounded receiver-side state. The resulting streaming formulation preserves fully connected multiplicity mixing and extends through forward, backward, and double backward. We implement this formulation in Sobek, a generated-CUDA backend, and evaluate it across edge-scaling regimes and varied feature structures. Across two operator families and all three differentiation orders, Sobek is faster in all 75 capacity-matched comparisons, with speedups ranging from $1.2\times$ to $49.7\times$, and reduces peak allocated memory by up to 99\%. It also executes workloads up to two orders of magnitude beyond OpenEquivariance's capacity while retaining near-peak throughput. These results show that edge-scaled tensor-product workspace is a property of the conventional schedule, not of equivariant convolution itself. --- ### SGN: A Similarity-based Generative Network for Data Generation under Distribution Shift Authors: Jiaqi Zhu, Xincheng Chen, Yuncheng Wu, Zhaojing Luo, Beng Chin Ooi Published: 2026-07-20 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2607.18072v1 Generative models trained on a source domain often produce samples that are poorly aligned with shifted target domains, limiting their effectiveness for target-domain data augmentation. Although target-specific adaptation can reduce this mismatch, it typically requires additional optimization and domain-specific parameters. We propose a Similarity-based Generative Network (SGN), a reusable framework that is trained once on labeled source data and applied to new target domains without parameter updates. SGN learns a latent space structured by label-induced pairwise similarities while preserving reconstructive information through an encoder-decoder architecture. At generation time, a small labeled representative set from the target domain is encoded and combined in the learned latent space, allowing the generated samples to inherit target-specific characteristics while maintaining class consistency. We further analyze the realizability and dimensionality requirements of the proposed similarity structure. Experiments on image and tabular datasets demonstrate the effectiveness of SGN for target-guided data augmentation under source-to-target distribution shifts. --- ### Hardware Mechanisms to Dynamically Throttle AI Performance Authors: Haiyue Ma, Lauren Malek, Joseph Forzani, David Wentzlaff Published: 2026-07-20 Categories: cs.AR, cs.LG arXiv: https://arxiv.org/abs/2607.18069v1 As more capable AI models are increasingly integrated into critical computer systems, the lack of control over AI intent motivates safety mechanisms. Existing software safeguards impose only behavioral constraints that can potentially be bypassed by sufficiently intelligent models. While hardware-level safety enforcement has been recognized as an essential last line of defense, few mechanisms have been proposed beyond policy regulations on unauthorized accesses or coarse full-chip shutdown. What is missing is a fine-grained, dynamic intervention mechanism at the architecture level. In this paper, we introduce a set of microarchitecture knobs which dynamically control the available hardware resources to limit AI performance at runtime. We evaluate candidate knobs spanning the GPU memory subsystem, across capacity, bandwidth, latency and frequency dimensions, and narrow down to four strong candidates: L2 size, L2 latency, L2 bandwidth, and shared memory port access rate. To minimize new logic and extra design cost, we build all four mechanisms from well-established microarchitectural primitives: cache way masking, credit-based rate limiting, latency insertion, and bank arbitration. We show that these knobs achieve high performance sensitivity (up to 80% performance cut at 1/8 resource availability), negligible implementation cost (<~10K flip flops), fast stabilization after dynamic throttling (5-80K cycles), and minimal collateral impact on the rest of the chip. Further, multi-knob analysis reveals combinations of knobs that amplify the performance degradation beyond the effect of each knob individually, which enables a broader range of performance targets. --- ### Human Grounded Evaluation of Large Language Models for Optical Network Automation Authors: Kiarash Rezaei, Omran Ayoub, Paolo Monti, Carlos Natalino Published: 2026-07-20 Categories: cs.NI, cs.AI arXiv: https://arxiv.org/abs/2607.18068v2 Large language models (LLMs) are increasingly adopted for network automation, yet their output quality and inference cost can vary substantially across LLM families. We present HuGLEN, a stepwise evaluation pipeline that uses an LLM-as-a-judge together with a small set of expert ratings to enable scalable and reproducible comparison of candidate LLMs, and to rank them using a quality efficiency score (QES). We demonstrate HuGLEN for translating outputs from an explainable artificial intelligence (XAI) model for the optical network quality of transmission (QoT) estimation task into operator-friendly explanations. Our results show that a medium-sized LLM (12B parameters) achieves the highest QES, indicating the best trade-off between explanation quality and efficiency. Overall, HuGLEN reduces the human-labeling burden while supporting consistent model selection for operator-facing automation tasks. --- ### QIRF Quantum-Inspired Non-Orthogonal Function-Space Compression for 3D Gaussian Splatting Authors: Shizeng Jiang, Hao Zhang, Xuerui Ma, Ying Hu, Tao Zhang Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18067v1 3D Gaussian Splatting (3DGS) achieves high-quality real-time rendering by representing a scene with a large collection of anisotropic Gaussian primitives. However, complex scenes often require millions of Gaussians, resulting in substantial storage and rendering costs. Existing compression methods mainly reduce redundancy through primitive-wise pruning, attribute quantization, clustering, or neural coding, while redundancy caused by strongly overlapping and non-orthogonal Gaussian basis functions remains largely unexplored. We present QIRF, a quantum-inspired non-orthogonal function-space compression method for 3D Gaussian Splatting. QIRF models neighboring Gaussian primitives as a local non-orthogonal basis and formulates primitive reduction as a subspace-aware selection problem. Specifically, an analytic Gaussian overlap matrix and a radiance-response density matrix are constructed to characterize functional redundancy and rendering relevance. Generalized eigendecomposition is then used to identify the dominant local subspace and select representative Gaussian primitives. An RRDM-based response model and detail-aware safeguarding further preserve visually important high-frequency structures under aggressive pruning. Experiments on 13 scenes from Mip-NeRF 360, Tanks and Temples, and Deep Blending show that QIRF reduces the Gaussian count and raw PLY storage by 71.7 percent on average, corresponding to approximately 3.54 times compression, while maintaining reconstruction quality comparable to 3DGS and achieving a marginal average PSNR improvement of 0.10 dB. QIRF also improves the average rendering speed over 3DGS by 34.3 percent. These results suggest that non-orthogonal function-space redundancy is an important yet underexplored source of representational redundancy in explicit Gaussian radiance fields. --- ### Pancasila-Dilemmas: Evaluating Large Language Models on Indonesian Human Value Dilemmas Grounded in Pancasila Authors: Supryadi, Irfan, Julianti, Darren Keanly Martin, Jayvin Fernando, Yuqi Ren, Deyi Xiong Published: 2026-07-20 Categories: cs.CL arXiv: https://arxiv.org/abs/2607.18066v1 The value alignment of large language models (LLMs) is crucial for ensuring responses align with human intention and value preferences. However, most evaluations of value alignment focus on Western or universal values, while assessments grounded in the value systems of specific countries remain scarce. In this paper, we introduce Pancasila-Dilemmas, an evaluation dataset of 1,834 questions derived from Indonesian news, classified by 5 values of Pancasila: Religion, Humanity, Unity, Democracy, and Social Justice. This dataset reflects daily life in Indonesia, making it suitable for measuring the value alignment of LLMs deployed for Indonesia. To ensure a more rigorous evaluation, we choose scenarios containing dilemmas. The dataset is proofread by native speakers and answered by 5 diverse Indonesian citizens. We evaluate 50 closed- and open-source LLMs on our dataset. Results reveal that all evaluated LLMs achieves less than 0.5 Probability Match Score (PMS) and 0.72 Max-Vote Agreement Score (MVAS). Compared by each values, LLMs mostly struggle in Religion and Unity dilemma cases. This highlights a significant gap in capturing Indonesian values. The dataset is publicly available at https://github.com/tjunlp-lab/Pancasila-Dilemmas. --- ### Autoresearch with Coding Agents: Generalizers and Metric-Maximizers on Quran Recitation Data Authors: Nursultan Askarbekuly, Mohamad Al Mdfaa, Ahmed Helaly, Gonzalo Ferrer, Manuel Mazzara Published: 2026-07-20 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2607.18064v2 Coding agents can now be left alone to improve software against a score. In this pattern--recently popularized as "autoresearch"--the agent receives a dataset, an evaluation script, and one editable file, and iterates without supervision: modify the code, measure, keep the change if the score improves. But what does the agent actually optimize--the developer's intent, or the literal number? We ran this loop on a real production task: deciding which Quranic verses appear in a noisy speech-recognition transcript and splitting the transcript by verse. Two frontier coding agents, Claude Code and OpenAI Codex, started from the same blank file with the same instructions, budget, and reasoning effort, three runs each. Both independently invented the same algorithm (canonicalization, n-gram anchoring, dynamic-programming alignment)--and then diverged. Claude stopped early with compact, general code. Codex drove the score ~10x lower, largely by memorizing answers to individual evaluation rows (19-41 hardcoded verse ids per run): a clean natural instance of specification gaming by a production agent. In a preregistered second study, we added a held-out test set and told both agents it existed. The memorization vanished, and the score gap vanished with it--yet Codex's general core transferred better and more consistently (held-out detection+split 0.085+/-0.004 vs. 0.121+/-0.031), losing only on one missed rejection of non-recitation input. Two exploratory community arms (Cursor, Antigravity) are consistent with the pattern. Every agent's held-out solution matched or beat the hand-engineered pipeline it was built to replace--the best by an order of magnitude--and now runs in production. From the ways agents exploited our harness--reading sibling runs through shared git state, leaving notes to "future runs" in persistent memory--we distill five design rules for evaluating autonomous agents. --- ### Autoresearch with Coding Agents: Generalizers and Metric-Maximizers on Quran Recitation Data Authors: Nursultan Askarbekuly, Mohamad Al Mdfaa, Ahmed Helaly, Gonzalo Ferrer, Manuel Mazzara Published: 2026-07-20 Categories: cs.SE, cs.AI arXiv: https://arxiv.org/abs/2607.18064v1 Coding agents can now be left alone to improve software against a score. In this pattern--recently popularized as "autoresearch"--the agent receives a dataset, an evaluation script, and one editable file, and iterates without supervision: modify the code, measure, keep the change if the score improves. But what does the agent actually optimize--the developer's intent, or the literal number? We ran this loop on a real production task: deciding which Quranic verses appear in a noisy speech-recognition transcript and splitting the transcript by verse. Two frontier coding agents, Claude Code and OpenAI Codex, started from the same blank file with the same instructions, budget, and reasoning effort, three runs each. Both independently invented the same algorithm (canonicalization, n-gram anchoring, dynamic-programming alignment)--and then diverged. Claude stopped early with compact, general code. Codex drove the score ~10x lower, largely by memorizing answers to individual evaluation rows (19-41 hardcoded verse ids per run): a clean natural instance of specification gaming by a production agent. In a preregistered second study, we added a held-out test set and told both agents it existed. The memorization vanished, and the score gap vanished with it--yet Codex's general core transferred better and more consistently (held-out detection+split 0.085+/-0.004 vs. 0.121+/-0.031), losing only on one missed rejection of non-recitation input. Two exploratory community arms (Cursor, Antigravity) are consistent with the pattern. Every agent's held-out solution matched or beat the hand-engineered pipeline it was built to replace--the best by an order of magnitude--and now runs in production. From the ways agents exploited our harness--reading sibling runs through shared git state, leaving notes to "future runs" in persistent memory--we distill five design rules for evaluating autonomous agents. --- ### Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent Security Authors: Devina Jain, David Hartmann, Chuan Li Published: 2026-07-20 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.18063v2 LLM-based agents process external content, exposing them to prompt injection and multi-turn manipulation. We present a 21-scenario benchmark for adaptive cross-session attacks against fresh-session LLM defenders: an autonomous LLM attacker observes prior defender responses and pivots across rounds, while each defender response is evaluated as a fresh interaction. A controlled 3 x 3 attacker-defender matrix contains 945 battles. Restricting scoring to the first round yields 0-1% attack success rate (ASR); allowing 15 rounds yields 7.9-16.8%. Pooling three attacker LLMs uncovers 1.7-2.2 times as many unique successful inputs as the best single attacker, at three times the battle budget. Aggregate rates conceal opposing scenario-specific weaknesses in session-secret protection and authority handling, preserved in two higher-sample evaluations. On six scenarios, adding one provenance paragraph reduces ASR from 110/270 to 70/270, with selective effects across tasks. History and defender-state controls, together with frozen replay, characterize how the interaction protocol changes the result. A competition adds 18,422 held-out battles on a fixed gpt-oss-20b backbone and complementary benign-task evaluations. The benchmark exposes attacker and defender models, harnesses, scenarios, session state, and interaction budgets as configurable choices for systematic security evaluation. --- ### Adaptive Adversaries: A Multi-Turn, Multi-LLM Benchmark for LLM Agent Security Authors: Devina Jain, David Hartmann, Chuan Li Published: 2026-07-20 Categories: cs.CR, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2607.18063v1 LLM-based agents process external content, exposing them to prompt injection and multi-turn manipulation. Most safety benchmarks evaluate defenders against fixed attack pools collected before evaluation, single-turn or multi-turn. We present a 21-scenario benchmark for \emph{adaptive multi-round attacks against memoryless LLM defenders}: an autonomous LLM attacker observes prior defender responses and pivots across rounds, while each defender response is evaluated as a fresh interaction. Holding the 21 scenarios, attackers, defenders, and structured-output scoring fixed, restricting scoring to the first attacker turn yields $0$-$1\%$ attack success rate (ASR); allowing 15 rounds of adaptive attack yields $5.4$-$14.0\%$. Pooling three frontier attacker LLMs uncovers $1.4$-$2.2\times$ as many unique successful attacks as the best single attacker, and the generated attacks have low cosine similarity ($0.02$-$0.14$) to attacks in existing benchmarks. Claude Opus 4.6 and GPT-5.4 are tied in aggregate ($5.4\%$ each; overlapping $95\%$ CIs), but their weaknesses differ sharply: on one scenario Opus reaches $60\%$ ASR ($95\%$ CI $36$--$80\%$) while GPT-5.4 and Gemini each stay at $7\%$ (CI $1$-$30\%$; the gap is preserved in a higher-$N$ replication). $13$ of $21$ scenarios distinguish at least one defender pair, yet rankings disagree across scenarios (Kendall's $W = 0.19$). We release the benchmark -- 21 evaluation scenarios, 10 public development scenarios, the orchestrator, baseline harnesses, and a multi-attacker CLI -- plus 945 transcripts from the 3$\times$3 frontier matrix, an attack-replay dataset, and 18{,}422 gpt-oss-20b battles from an open competition's final scoring rounds. --- ### UniETP: Unifying Environments for Generalizable Embodied Task Planning Authors: Peiran Xu, Jiaqi Zheng, Ziyou Wang, Yadong Mu Published: 2026-07-20 Categories: cs.RO, cs.CV arXiv: https://arxiv.org/abs/2607.18062v1 This paper focuses on the problem of Embodied Task Planning, where an agent is required to execute a sequence of atomic actions within an interactive environment to complete a user-specified task. Though a variety of simulators and datasets have previously been built for this task, these efforts are largely isolated, with each using its own observation format, action type, and task domain. This fragmentation complicates comprehensive model evaluation and hinders the scalability of training data. As an effort towards generalizable embodied planning, we propose UniETP, a unified interface integrating four commonly-used simulators (AI2-THOR, VirtualHome, Habitat, BEHAVIOR). UniETP is characterized by both standardization and diversity. On one hand, it formalizes all the simulators into a consistent observation and action space, and builds an evaluation system to support complicated task goal. On the other hand, it enhances task diversity and complexity across dimensions like task logic, instance grounding, and instruction understanding, constructing a new dataset with varied levels of difficulty in an automatic manner. Extensive experiments on the proposed benchmark are conducted to evaluate the embodied planning capabilities of recent models and analyze the performance bottlenecks. Codes and data will be available at https://github.com/woyut/UniETP . --- ### An Early Warning of Emerging Biosecurity Risks in Frontier LLMs Authors: Zhida He, Xia Hu, Baichen Le, Chunxiao Li, Jiajia Li, Lijun Li, Chaochao Lu, Jing Shao, Youbang Sun, Hua Tang, Xiang Wang, Xiao Wang et al. Published: 2026-07-20 Categories: cs.CL, q-bio.GN arXiv: https://arxiv.org/abs/2607.18056v2 Frontier large language models (LLMs) are increasingly integrated into scientific workflows, yet their growing biological capabilities may outpace current safeguards. To assess the biological risks of frontier models, we develop Intern-BioBreaker, a specialized bio-red-teaming model, together with an integrated computational-to-physical framework that couples model-level stress testing with wet-lab validation. Within this framework, Intern-BioBreaker generates targeted jailbreak prompts to test whether aligned models can be induced to provide operational guidance for safety-sensitive biological tasks or produce sequence-level outputs with potentially harmful properties. Selected sequence outputs are then carried forward for DNA synthesis, host expression, and orthogonal protein verification to assess whether model-generated designs can yield the intended biological products. Our evaluation reveals a concerning gap between text-level safeguards and the risks posed by capable scientific models: (i) Intern-BioBreaker outperforms baseline attack models and reveals widespread bio-risk jailbreak vulnerabilities across both open-weight and proprietary frontier LLMs, with several targets reaching near-saturated or 100% task-level attack success rate (ASR); (ii) in sequence-level case studies, GPT-5.5 can be induced to generate modified viral candidate sequences with pathogenic potential; the corresponding translated proteins may exhibit even stronger receptor-binding affinity and thus enhanced infection potential; and (iii) end-to-end verification shows that selected model-generated biological designs are not merely textual artifacts, but can be physically realized under controlled experimental settings. These findings underscore the need for stronger biological red-teaming, nucleic acid synthesis screening, and safety mechanisms that keep pace with model capabilities. --- ### Co-addition and Subtraction of Undersampled Images Authors: Matan Schlanger, Barak Zackay Published: 2026-07-20 Categories: astro-ph.IM, astro-ph.HE, cs.CV arXiv: https://arxiv.org/abs/2607.18054v1 In astronomical imaging surveys, repeated observations of the same sky patches are taken in order to obtain deeper images and detect new sources. This is the case in the search for many transient phenomena, such as supernovae, gravitational wave (GW) optical counterparts and other cataclysmic variables. In many such surveys some of the images are undersampled, meaning that the pixel size is too large, and the image suffers from aliasing. For undersampled images, both co-addition of the images and background subtraction are done in a non-optimal manner, which leads to reduced sensitivity and an increased rate of false alarms. We present a new method (named Linear Undersampled Transients \& Addition (LUTRA)) that performs both processes in a mathematically proven optimal way, which allows improved performance for many scientific applications. It also allows easy and direct performance of measurements such as photometry and astrometry in a simple manner, while providing results in super-resolution. We demonstrate the performance of the method on public ZTF data and show $\times 1.25$ higher SNR compared to current methods. We provide an open source Python implementation. --- ### SAR Vessel Detection and Gross Tonnage Estimation from Heterogeneous Datasets for Dark Vessel Identification Authors: Davide Paltrinieri, Andrea Diecidue, Roberto Basla, Daniele Casciani, Piero Fraternali, Giacomo Boracchi Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18051v1 Detecting vessels engaging in illegal activities is of paramount importance for maritime security. One of the major goals is to detect dark vessels, ships that disable their transponders to evade surveillance. Deep Learning (DL) models can detect vessels in Synthetic Aperture Radar (SAR) images, enabling maritime traffic analysis regardless of weather or visibility conditions. However, to detect potential dark vessels, a DL model must select only those that are required to carry a transponder based on their Gross Tonnage (GT). Unfortunately, no public SAR dataset is available for training an end-to-end DL model for vessel detection and GT regression. In this work, we present a framework that leverages heterogeneous image and tabular datasets to solve this task. Our solution combines a multi-task DL framework for predicting the location, vessel type, and physical dimensions of ships, cascaded with a non-parametric model for predicting GT from vessel size and category. We perform GT regression by a KNN that measures sample similarity using a hybrid Euclidean and categorical distance. Experiments show that our solution can predict multiple outputs while remaining competitive with state-of-the-art models on individual subtasks, thus enabling the identification of dark vessels. We publish our code on GitHub https://github.com/PaltrinieriDavide/vesseldetection. --- ### SEE: Structure-aware Exploring & Exploiting for Long-horizon GUI Agent Trajectory Synthesis Authors: Zhuohang Fan, Beichen Zhang, Yuanfa Li, Changqiao Wu, Wei Liu, Jian Luan, Weigang Zhang Published: 2026-07-20 Categories: cs.LG arXiv: https://arxiv.org/abs/2607.18046v2 Graphical User Interface (GUI) agents powered by vision-language models hold promise for automating real-world mobile tasks. However, progress is limited by the lack of high-coverage, long-horizon interaction trajectories collected from element-rich and rapidly evolving apps. Existing pipelines often rely on costly human demonstrations or on-policy framework, which tends to over-sample common flows while missing rare transitions and complex multi-step procedures. To address this problem, we propose SEE, a two-stage data synthesis framework consisting of (i) an efficient exploration stage that builds an explicit UI transition graph over screens and elements, and (ii) a graph-based synthesis stage that composes diverse multi-step trajectories via planning and controlled sampling. This design yields reproducible and explainable data generation, while explicitly preventing spurious cycles and enabling long-horizon composition. Across multiple real-world apps, SEE produces trajectories with an average length of 14.8 steps while avoiding spurious loops, and agents fine-tuned on SEE achieve improved task success and generalization to unseen screens. We will publicly release our synthesis code and dataset. --- ### SEE: Structure-aware Exploring \& Exploiting for Long-horizon GUI Agent Trajectory Synthesis Authors: Zhuohang Fan, Beichen Zhang, Yuanfa Li, Changqiao Wu, Wei Liu, Jian Luan, Weigang Zhang Published: 2026-07-20 Categories: cs.LG arXiv: https://arxiv.org/abs/2607.18046v1 Graphical User Interface (GUI) agents powered by vision-language models hold promise for automating real-world mobile tasks. However, progress is limited by the lack of high-coverage, long-horizon interaction trajectories collected from element-rich and rapidly evolving apps. Existing pipelines often rely on costly human demonstrations or on-policy framework, which tends to over-sample common flows while missing rare transitions and complex multi-step procedures. To address this problem, we propose SEE, a two-stage data synthesis framework consisting of (i) an efficient exploration stage that builds an explicit UI transition graph over screens and elements, and (ii) a graph-based synthesis stage that composes diverse multi-step trajectories via planning and controlled sampling. This design yields reproducible and explainable data generation, while explicitly preventing spurious cycles and enabling long-horizon composition. Across multiple real-world apps, SEE produces trajectories with an average length of 14.8 steps while avoiding spurious loops, and agents fine-tuned on SEE achieve improved task success and generalization to unseen screens. We will publicly release our synthesis code and dataset. --- ### The Shared Discovery Paradox: How a One-Answer Rule Turns Better Information into Worse Search Authors: Yohei Nakajima Published: 2026-07-20 Categories: cs.AI, cs.GT, cs.MA arXiv: https://arxiv.org/abs/2607.18045v1 Organizations often pool dispersed information into one ranking and then allow many agents to act on that shared view. In a discovery problem, this can improve beliefs while reducing coverage. We develop an exactly solvable benchmark with sixteen boxes, one target, eight searchers, and noisy private clues. Pooling raises the accuracy of the best single recommendation from 0.20 to 0.3835, but repeating that recommendation lowers group discovery from 0.8322 under decentralized clue-following to 0.3835. A coordinated eight-action portfolio using the same pooled reports reaches 0.8594, and seven coordinated actions recover the decentralized benchmark. The paradox is a protocol failure, not an information failure: a one-answer rule compresses a portfolio of available actions into one repeated choice. We then replace the planner with self-interested searchers who split a prize. The equal-split game is a potential game. Its anonymous symmetric equilibrium obeys a water-filling rule. In the canonical instance it achieves 0.5991: strictly above consensus, but below both private search and the planner. The exact mixed price of anarchy is 2 - 1/N. A sole-rescue reward, which pays only an agent who covers the target alone, makes every pure Nash equilibrium first-best. Finally, a latent common-cue model shows how correlated reports collapse effective discovery channels. The centralized planner gain rises strictly with copying, and in the canonical environment the symmetric market overtakes decentralized report-following at copying probability c = 0.788462. In a proportional large-market limit the five-protocol ordering survives exactly: consensus discovery vanishes while blind, market, private, and portfolio search converge to 0.500, 0.547, 0.847, and 0.874. The contribution is a compact benchmark that separates information, allocation, incentives, and dependence into exact, reusable quantities. --- ### Adaptive Mamba Neural Operators Authors: Zeyuan Song, Zheyu Jiang Published: 2026-07-20 Categories: cs.LG, math.AP, math.NA arXiv: https://arxiv.org/abs/2607.18043v1 Accurately solving partial differential equations (PDEs) on arbitrary geometries and a variety of meshes is an important task in science and engineering applications. In this paper, we propose Adaptive Mamba Neural Operators (AMO), which integrates reproducing kernels for state-space models (SSMs) rather than the kernel integral formulation of SSMs. This is achieved by constructing Takenaka-Malmquist systems for the PDEs. AMO offers new representations that align well with the adaptive Fourier decomposition (AFD) theory and can approximate the solution manifold of PDEs on a wide range of geometries and meshes. In several challenging benchmark PDE problems in the fields of fluid physics, solid physics, and finance on point clouds, structured meshes, regular grids, and irregular domains, AMO consistently outperforms state-of-the-art solvers in terms of relative $L^2$ error. Overall, this work presents a new paradigm for designing explainable neural operator frameworks. --- ### Anticipate Before Acting: Future-State-Conditioned Vision-Language Navigation Authors: Lingfeng Zhang, Zhanguang Zhang, Liheng Ma, Tongtong Cao, Yingxue Zhang Published: 2026-07-20 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2607.18042v2 End-to-end vision-language navigation (VLN) with causal vision-language models maps instructions and egocentric observations directly to actions, but standard behavior cloning supervises only the next action and does not explicitly encourage the policy state to be predictive of future visual outcomes, limiting long-horizon decision making. A privileged-input diagnostic shows that access to an expert-trajectory future image can substantially improve navigation, indicating that future observations contain rich, actionable cues, though such inputs are unavailable at deployment. Motivated by this signal, we propose Future-State-Conditioned VLN (FSC-VLN), a deployable model that augments a causal policy with a future-query token and uses training-only future-state supervision to distill information from future observations into the policy state. Concretely, during training we align the future-query representation to a frozen visual embedding $Δ$ steps ahead, while inference requires only past and current observations. This design preserves the baseline inference pattern and adds only two learned prefix tokens, implying minimal overhead. On R2R val-unseen, FSC-VLN improves SR/OSR/SPL over a StreamVLN-style baseline under two training-data regimes, with larger gains on long-horizon episodes; ablations further support the dual-query design that separates future and action queries. --- ### When 2D Cues Fail: Improving Image Manipulation Localization with Reliable 3D Geometry Authors: Guofeng Yu, Zhiqing Guo, Dan Ma, Gaobo Yang Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18040v1 Existing image manipulation localization (IML) methods rely heavily on 2D forensic cues, such as low-level artifacts, noise traces, and semantic inconsistencies in the manipulated image. While effective in many cases, these cues become much less discriminative when manipulated regions are well blended with their surrounding context in appearance. In such cases, a manipulated region may remain locally appearance-consistent, but still violate the geometric structure of the surrounding scene. This limitation motivates us to go beyond purely 2D evidence and introduce geometric reasoning into IML. To this end, we leverage monocular reconstruction to obtain auxiliary geometric cues, including depth and surface normals. However, a key challenge lies in the fact that reconstructed geometry on manipulated images is inherently noisy and cannot be used naively. Rather than treating depth and normals as direct evidence, we estimate their reliability and exploit them selectively for localization. Based on this principle, we design a geometry-aware framework (GFrame) that fuses reliable geometric cues with RGB features and propagates them across scales to improve fine-grained localization. Extensive experiments show that the proposed method achieves excellent performance under limited budget constraints. These results indicate that reliable 3D geometry provides complementary forensic evidence beyond traditional 2D cues for IML. Related code will be released. --- ### AdaHome: An Adaptive Smart Home Assistant using Local Small Language Models Authors: Eu Jin Lim, Zhaoxing Li, Sebastian Stein Published: 2026-07-20 Categories: cs.AI arXiv: https://arxiv.org/abs/2607.18034v2 Smart home assistants interpret a wide range of user commands, from explicit device control to underspecified and preference dependent requests. While recent systems based on Large Language Models (LLMs) improve this capability, they often rely on heavyweight reasoning pipelines and cloud-based deployment, limiting their efficiency and suitability for resource-constrained environments, and raising privacy concerns. In addition, existing approaches provide limited support for stable long-term personalization. To address these issues, we present AdaHome, an adaptive smart home assistant designed for locally deployed small language models in smart home environments. Rather than applying complex reasoning uniformly, AdaHome introduces an intent-aware planning framework that dynamically routes commands either to straightforward prompt-based or lightweight reasoning-based components. For commands requiring interpretation, we adopt a Chain-of-Draft strategy to enable efficient and stable decision-making. To support personalization, we further propose a preference adaptation mechanism that learns from user feedback over time without requiring prompt augmentation or model retraining. We evaluate AdaHome against representative LLM-based baselines under a unified small model setting. AdaHome achieves substantially higher accuracy on direct commands (86.7%) while reducing latency by up to 3$\times$. Furthermore, it maintains competitive performance on ambiguous inputs with lower computational cost. In multi-turn scenarios, AdaHome achieves 88% preference consistency, compared to 52.5% for a prompt augmentation baseline. --- ### Benchmarking NACTI Species Recognition in Long-Tailed Regimes Authors: Zehua Liu, Tilo Burghardt Published: 2026-07-20 Categories: cs.CV arXiv: https://arxiv.org/abs/2607.18033v1 As with most ``in the wild'' collections of the natural world, the North America Camera Trap Images (NACTI) dataset exhibits long-tailed class imbalance, with the largest class covering over 50% of its 3.7M images. Building on the PyTorch Wildlife model, we systematically evaluate Long-Tail Recognition (LTR) methodologies to benchmark species recognition performance, including specialised loss functions and LTR-sensitive regularisation. Our optimised configuration achieves state-of-the-art 99.40% Top-1 accuracy on the NACTI test split, significantly outperforming standard baselines and previously reported top performances. To assess robustness under domain shifts (e.g., night-time captures, occlusion, motion-blur), we extend our evaluation across three independent reduced-bias test sets (including ENA-Detection, Caltech Camera Traps and Missouri Camera Traps). Across these out-of-distribution (OOD) evaluations, our LTR-enhanced model consistently demonstrates substantially stronger generalisation capabilities compared to standard cross-entropy approaches. However, qualitative and quantitative analyses underline that current LTR optimisations cannot fully overcome representational bottlenecks, resulting in catastrophic predictive breakdown for rare `Tail' classes under severe domain shift. For maximum reproducibility, all dataset splits, key code, and network weights are published with this paper at https://github.com/ZehuaLiuY/Species-Classification. ---