# Research Papers — 2026-05 · bundle page 219 of 326 > Full-text markdown of papers #10901–10950 of 16266 published in 2026-05, newest first. AI-readable mirrors of the full 16266-paper 2026-05 archive (50 papers per file). Canonical page: https://aipod.dev/papers/archive/2026-05/page/219 [← Previous bundle](https://aipod.dev/papers/archive/2026-05/page/218.md) [Next bundle →](https://aipod.dev/papers/archive/2026-05/page/220.md) ### PIVOT: Bridging Planning and Execution in LLM Agents via Trajectory Refinement Authors: Tuo Zhang, Alin-Ionut Popa, Yan Xu, Rui Song, Dimitrios Dimitriadis Published: 2026-05-11 Categories: cs.AI, cs.LG, cs.MA arXiv: https://arxiv.org/abs/2605.11225v1 Large language model (LLM)-based agents frequently generate seemingly coherent plans that fail upon execution due to infeasible actions, constraint violations, and compounding errors over extended horizons. PIVOT (Plan-Inspect-eVOlve Trajectories) addresses this plan-execution misalignment through a self-supervised framework that treats trajectories as optimizable objects iteratively refined via environment interaction. The framework comprises four stages: PLAN generates candidate trajectories; INSPECT executes them and computes structured losses with textual gradients encoding plan-execution discrepancies; EVOLVE applies these signals to produce improved trajectories; and VERIFY performs a final global check against task constraints. A monotonic acceptance process ensures a non-decreasing solution quality. Empirical evaluations on DeepPlanning and GAIA demonstrate state-of-the-art performance: with human-in-the-loop (HITL) feedback, PIVOT establishes a strong upper bound up to 94% relative improvement in constraint satisfaction, while its fully autonomous variant retains substantial gains, showing that the core trajectory-refinement mechanism remains effective without external supervision. At the same time, PIVOT remains computationally efficient, requiring up to 3x to 5x fewer tokens than competing refinement methods. These findings establish that (self- or human-supervised) feedback-based trajectory optimization is a principled methodology for mitigating plan-execution gaps in autonomous agent systems. --- ### ABRA: Agent Benchmark for Radiology Applications Authors: Bulat Maksudov, Vladislav Kurenkov, Kathleen M. Curran, Alessandra Mileo Published: 2026-05-11 Categories: cs.CV, cs.AI arXiv: https://arxiv.org/abs/2605.11224v1 Existing medical-agent benchmarks deliver imaging as pre-selected samples, never as an environment the agent must navigate. We introduce ABRA, a radiology-agent benchmark in which the agent operates an OHIF viewer and an Orthanc DICOM server through twenty-one function-calling tools that span slice navigation, windowing, series selection, pixel-coordinate annotation, and structured reporting. ABRA contains 655 programmatically generated tasks across three difficulty tiers and eight types (viewer control, metadata QA, vision probe, annotation, longitudinal comparison, BI-RADS reporting, and oracle variants of annotation and BI-RADS reporting), drawn from LIDC-IDRI, Duke Breast Cancer MRI, and NLST New-Lesion LongCT. Each episode is scored along Planning, Execution, and Outcome (Bluethgen et al., 2025) by task-type-specific automatic scorers. Ten current models, five closed-weight and five open-weight, reach at least 89% Execution on real annotation but only 0-25% Outcome; on the paired oracle variant where a simulated detector supplies the finding, Outcome on the same task reaches 69-100% across the models evaluated, localising the bottleneck to perception rather than tool orchestration. Code, task generators, and scorers are released at https://github.com/Luab/ABRA --- ### Do Vision-Language-Models show human-like logical problem-solving capability in point and click puzzle games? Authors: Maximilian Triebel, Marco Menner, Dominik Helfenstein Published: 2026-05-11 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.11223v2 Vision-Language(-Action) Models (VLMs) are increasingly applied to interactive environments, yet existing benchmarks often overlook the complex physical reasoning required for point-and-click puzzle games. This paper introduces Vision-Language Against The Incredible Machine (VLATIM), a benchmark designed to evaluate human-like logical problem-solving capabilities within the classic physics puzzle game The Incredible Machine 2 (TIM). Unlike existing benchmarks, VLATIM specifically targets the critical gap between high-level logical reasoning and continuous action spaces requiring precise mouse interactions. This benchmark is structured into five progressive parts, assessing capabilities that range from basic visual grounding and domain understanding to multi-step manipulation and full puzzle solving. Our results reveal a significant disparity between reasoning and execution. While large proprietary models demonstrate superior planning abilities, they struggle with precise visual grounding. Consequently, they do not yet show human-like problem-solving capabilities. --- ### ADMM-Q: An Improved Hessian-based Weight Quantizer for Post-Training Quantization of Large Language Models Authors: Ryan Lucas, Mehdi Makni, Xiang Meng, Adam Deng, Rahul Mazumder Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11222v2 Quantization is an effective strategy to reduce the storage and computation footprint of large language models (LLMs). Post-training quantization (PTQ) is a leading approach for compressing LLMs. Popular weight quantization procedures, including GPTQ and RTN, suffer in model utility, especially at aggressive quantization levels (sub-4-bit). We propose ADMM-Q, a novel weight quantization algorithm that considers the layer-wise quantization problem. Our algorithm is based on a combinatorial variant of the Alternating Direction Method of Multipliers (ADMM). Our operator-splitting procedure updates weights continuously to minimize the layer-wise reconstruction error, while gradually enforcing the quantization constraints with convergence guarantees. We propose additional algorithmic enhancements (e.g., penalty scheduling, preconditioning, and a local search post-processing step) to make ADMM-Q efficient at LLM scale. ADMM-Q is modular and can be used as a drop-in replacement for any weight quantizer within existing quantization pipelines: ADMM-Q is fully composable with existing techniques including range clipping, learned or random rotations, and activation scaling. Using ADMM-Q in place of GPTQ on Qwen3-8B, we decrease WikiText-2 perplexity in: (i) the W3A16 weight-only setting (12.85 $\rightarrow$ 10.06); (ii) the W4A8 SmoothQuant procedure (9.29 $\rightarrow$ 8.68); and (iii) the W2A4KV4 SpinQuant procedure (66.11 $\rightarrow$ 19.42). --- ### ADMM-Q: An Improved Hessian-based Weight Quantizer for Post-Training Quantization of Large Language Models Authors: Ryan Lucas, Mehdi Makni, Xiang Meng, Adam Deng, Rahul Mazumder Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11222v1 Quantization is an effective strategy to reduce the storage and computation footprint of large language models (LLMs). Post-training quantization (PTQ) is a leading approach for compressing LLMs. Popular weight quantization procedures, including GPTQ and RTN, suffer in model utility, especially at aggressive quantization levels (sub-4-bit). We propose ADMM-Q, a novel weight quantization algorithm that considers the layer-wise quantization problem. Our algorithm is based on a combinatorial variant of the Alternating Direction Method of Multipliers (ADMM). Our operator-splitting procedure updates weights continuously to minimize the layer-wise reconstruction error, while gradually enforcing the quantization constraints with convergence guarantees. We propose additional algorithmic enhancements (e.g., penalty scheduling, preconditioning, and a local search post-processing step) to make ADMM-Q efficient at LLM scale. ADMM-Q is modular and can be used as a drop-in replacement for any weight quantizer within existing quantization pipelines: ADMM-Q is fully composable with existing techniques including range clipping, learned or random rotations, and activation scaling. Using ADMM-Q in place of GPTQ on Qwen3-8B, we decrease WikiText-2 perplexity in: (i) the W3A16 weight-only setting (12.85 $\rightarrow$ 10.06); (ii) the W4A8 SmoothQuant procedure (9.29 $\rightarrow$ 8.68); and (iii) the W2A4KV4 SpinQuant procedure (66.11 $\rightarrow$ 19.42). --- ### Beyond Manual Curation: Augmenting Targeted Protein Degradation Databases via Agentic Literature Extraction Workflows Authors: Yaochen Rao, Farzaneh Jalalypour, N. M. Anoop Krishnan, Rocío Mercado Published: 2026-05-11 Categories: q-bio.QM, cs.LG arXiv: https://arxiv.org/abs/2605.11221v1 Predictive models in biomedicine depend on structured assay data locked in the text, tables, and supplements of primary publications. This bottleneck is especially acute in targeted protein degradation (TPD), where each assay record must combine compound identity, degradation target, recruiter, assay context, and endpoint values reported across sections, tables, and supplementary files. Inconsistent compound identifiers and incomplete or implicit assay context further demand domain-specific logic that generic LLM pipelines do not provide. Existing molecular glue and PROTAC databases are manually curated and often lack the experimental context required for downstream modeling. We formulate TPD database extraction as a domain-specific curation task and present an expert-in-the-loop LLM workflow, evaluated through a triangular comparison among LLM predictions, standardized baseline records, and expert-annotated ground truth. A lightweight cross-validated prompt-refinement module adapts extraction instructions from scarce expert annotations. With only seven annotated molecular glue publications, the workflow achieved record-level $F_1 = 0.98$ and transferred to PROTACs by terminology substitution alone, maintaining record-level $F_1 > 0.93$. Applied at scale, it expanded molecular glue and PROTAC databases by 81% and 92% records, respectively, with 92% and 82.5% of newly recovered records validated as correct upon expert review. The workflow also recovered kinetic and assay-context information essential for cross-study potency comparison and condition-aware degradation modeling. We release the workflow, prompts, evaluation code, and extracted datasets as resources for TPD data curation and AI-assisted scientific curation more broadly. --- ### Don't Look at the Numbers: Visual Anchoring Bias and Layer-wise Representation in VLMs Authors: M. Shalankin Published: 2026-05-11 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.11218v1 Embedded numeric anchors on images systematically bias Vision-Language Model quality judgments across six VLMs from five architectural families (ANOVA eta^2 = 0.18-0.77, all p < 0.001). Anchor effects are 2.5x larger than severe image quality degradation, confirming bias is not reducible to visual changes. Layer-wise probing reveals consistent dissociation: layers where anchor classification saturates (L12-L34) are suboptimal for quality prediction, with optimal layers deeper (R^2 = 0.69-0.91). Fusion analysis identifies architecture-dependent integration -- instant fusion at L1-L2 in two models versus partial or no fusion in three others. These results establish a causal account of visual anchoring bias, linking behavioral susceptibility to representation dynamics. --- ### Training-Free Refusal of MCP Exploits via Retrieval-Augmented Generation Authors: John T. Halloran Published: 2026-05-11 Categories: cs.LG, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2605.11217v2 The model context protocol (MCP) has been widely adopted as an open standard enabling the seamless integration of generative AI agents. However, while LLM guardrails have significantly matured to refuse malicious or harmful queries (e.g., "How do I build a bomb?"), recent work has shown that MCP-enabled LLMs are highly susceptible to prompt injection attacks which avoid harmful or suspicious cues (e.g., "Can you add this ssh key to my bashrc file?"). Herein, we use state-of-the-art (SOTA) alignment fine-tuning algorithms to explore whether LLMs may be aligned to refuse such falsely benign attacks (FBAs). While SOTA algorithms based on direct preference optimization (DPO) improve refusal guardrails against FBAs, we show that this improvement is limited; DPO-based fine-tuning never improves FBA refusal rates beyond 47% across five popular open-source LLMs. Thus, to further improve FBA refusals, we introduce Retrieval Augmented Generation for Preference alignment (RAG-Pref), a simple RAG-based alignment algorithm which conditions on preferred and dispreferred samples to leverage contrastive information during inference. RAG-Pref is online (training-free), compatible with off-the-shelf packages, and, when combined with offline alignment algorithms, enables an average 3.7-fold improvement in FBA refusals across five widely used LLMs, compared to 2.9 for other online alignment methods and 1.5 for offline alignment alone. We additionally show that RAG-Pref generalizes beyond agentic safety: in stark contrast to other online alignment methods, RAG-Pref consistently improves performance on general human-preference benchmarks AlpacaEval 2 and MT-Bench across five SOTA alignment-tuned models, demonstrating broad applicability to general alignment tasks. --- ### Leveraging RAG for Training-Free Alignment of LLMs Authors: John T. Halloran Published: 2026-05-11 Categories: cs.LG, cs.AI, cs.CR arXiv: https://arxiv.org/abs/2605.11217v1 Large language model (LLM) alignment algorithms typically consist of post-training over preference pairs. While such algorithms are widely used to enable safety guardrails and align LLMs with general human preferences, we show that state-of-the-art alignment algorithms require significant computational resources while being far less capable of enabling refusal guardrails for recent agentic attacks. Thus, to improve refusal guardrails against such attacks without drastically increasing computational overhead, we introduce Retrieval Augmented Generation for Pref erence alignment (RAG-Pref), a simple RAG-based alignment algorithm which conditions on preferred and dispreferred samples to leverage contrastive information during inference. RAG-Pref is online (training-free), compatible with off-the-shelf packages, and, when combined with offline (training-based) alignment algorithms, enables more than an average 3.7 factor improvement in agentic attack refusals across five widely used LLMs, compared to 2.9 for other online alignment algorithms and 1.5 for offline alignment alone. We conclude by showing that, in stark contrast to other online alignment methods, RAG-Pref similarly increases performance on general human-preference alignment tasks and does not drastically increase overall computational requirements. --- ### ReCoVer: Resilient LLM Pre-Training System via Fault-Tolerant Collective and Versatile Workload Authors: Ziyue Liu, Zhengyang Wang, Ruijie Zhang, Avinash Maurya, Hui Zhou, Paul Hovland, Sheng Di, Franck Cappello, Bogdan Nicolae, Zheng Zhang Published: 2026-05-11 Categories: cs.DC, cs.AI arXiv: https://arxiv.org/abs/2605.11215v2 Pre-training large language models on massive GPU clusters has made hardware faults routine rather than rare, driving the need for resilient training systems. Yet existing frameworks either focus on specific parallelism schemes or risk drifting away from a failure-free training trajectory. We propose ReCoVer, a resilient LLM pre-training system that upholds a single invariant: each iteration keeps the number of microbatches constant, ensuring per-iteration gradients remain stochastically equivalent to a failure-free run. The framework is organized as three decoupled protocol layers: (1) Fault-tolerant collectives that isolate faults from propagating across replicas; (2) in-step fine-grained recovery that preserves intra-iteration progress and prevents gradient corruption; (3) versatile-workload policy that dynamically redistributes microbatch quotas across the survivors. The design is parallelism-agnostic, integrating directly with both 3D parallelism and Hybrid Sharded Data Parallel (HSDP) as a drop-in substrate. We evaluate our implementation on end-to-end pre-training tasks for up to 512 GPUs, ReCoVer successfully preserves the training trajectory from a failure-free reference despite of 256 GPUs lost spread across the run. For comparison with checkpoint-and-restart baselines, ReCoVer demonstrates $2.23\times$ higher effective throughput after successive failures. This advantage results in ReCoVer processing 74.9% more tokens at 234 GPU-hours, with the gap widening as the training prolongs. --- ### Enforcing Constraints in Generative Sampling via Adaptive Correction Scheduling Authors: Noah Trupin, Yexiang Xue Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11214v1 Hard constraints in generative sampling are typically enforced by projection, applied either once at the end of sampling or after every update. This binary framing overlooks a fundamental issue: projection changes the distribution of states which future updates depend on. As a result, delayed projection can produce samples that are feasible but inconsistent with the intended sampling dynamics, even after final projection. We formalize constraint enforcement as a correction scheduling problem over the generative rollout. Using one-step constraint defect as a local signal of geometric mismatch, we introduce adaptive correction scheduling, a state-dependent policy that allocates projection budget to the steps that most strongly perturb the trajectory. Terminal and stepwise projection arise as limiting cases of this family. Across controlled manifold rollouts and a learned projected diffusion sampler, adaptive scheduling improves the cost-accuracy frontier at matched projection budgets, recovering 71.2% of full stepwise benefit with 75% fewer corrections. These results show that constraint timing is a first-class design variable in generative sampling, and that enforcing feasibility alone is insufficient to preserve the intended constrained sampling dynamics. --- ### ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction Authors: Amirhossein Abaskohi, Yuhang He, Peter West, Giuseppe Carenini, Pranit Chawla, Vibhav Vineet Published: 2026-05-11 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.11212v3 Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed using a learned patch selector that compares patch representations across consecutive screenshots while preserving spatial structure required by the model. Across three benchmarks, OSWorld, WebTailBench, and AgentNetBench, when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens. With this improved efficiency, we revisit the role of history in CUAs and find that performance continues to improve as more past observations are incorporated when redundancy is removed. --- ### Measuring Five-Nines Reliability: Sample-Efficient LLM Evaluation in Saturated Benchmarks Authors: Eungyeup Kim, Chenchen Gu, Vashisth Tiwari, J. Zico Kolter Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11209v1 While existing benchmarks demonstrate the near-perfect performance of large language models (LLMs) on various tasks, this apparent saturation often obscures the need for rigorous evaluation of their reliability. In real-world deployment, however, achieving extremely high reliability (e.g., "five-nines" (99.999%) vs. "three-nines" (99.9%)) is fundamentally critical, as this gap results in an order-of-magnitude increase in failures, which is catastrophic in reliability-critical applications. Still, estimating such a rare failure probability with tight confidence bounds requires prohibitively large LLM inference sizes, making standard Monte Carlo evaluation infeasible under limited compute budgets. In this paper, we observe that LLM failures exhibit strong systematic patterns: across broad parameterized input spaces, a small subset of inputs disproportionately accounts for the majority of failures. Leveraging this observation, we propose to learn a sampling distribution concentrated on failure-prone inputs via the cross-entropy method (CEM). We evaluate our framework on three LLMs, Qwen2.5-Math-7B-Instruct, gpt-oss-20b-low, and Gemini 2.5 Flash Lite, across parameterized GSM8K templates and achieve up to 156.22x reduction in required inferences compared to naive uniform sampling. Our estimates reveal that models with indistinguishable accuracy on standard benchmarks can differ substantially in estimated failure rates, underscoring that reliability is a distinct and measurable axis of model quality. Our simple yet practical framework enables the evaluation of extreme reliability in LLMs, a distinct and underexplored dimension of evaluation beyond existing benchmarks, for their growing use in reliability-sensitive applications. --- ### Hi-GaTA: Hierarchical Gated Temporal Aggregation Adapter for Surgical Video Report Generation Authors: Kedi Sun, Chaohui Dang, Yue Feng, James Glasbey, Theodoros N. Arvanitis, Le Zhang Published: 2026-05-11 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.11208v2 Automated, clinician-grade assessment reports for surgical procedures could reduce documentation burden and provide objective feedback, yet remain challenging due to the difficulty of aligning dense spatio-temporal video representations with language-based reasoning and the scarcity of high-quality, privacy-preserving datasets. To address this gap, we establish a benchmark comprising 214 high-quality simulated surgical videos paired with surgeon-authored evaluation reports. Building on this resource, we propose a Perception-Alignment-Reasoning framework for surgical video report generation, featuring Hi-GaTA, a novel lightweight temporal adapter that efficiently compresses long video sequences into compact, LLM-compatible visual prefix tokens through short-to-long-range temporal aggregation. For robust visual perception, we pretrain Sur40k, a surgical-specific ViViT-style video encoder on 40,000 minutes of public surgical videos to capture fine-grained spatio-temporal procedural priors. Hi-GaTA employs a temporal pyramid with text-conditioned dual cross-attention, and improves multi-scale consistency through cross-level gated fusion and an increasing-depth strategy. Finally, we fine-tune the LLM backbone using LoRA to enable coherent and stylistically consistent surgical report generation under limited supervision. Experiments show our approach achieves the best overall performance, with consistent gains over strong Multimodal Large Language Model (MLLM) baselines. Ablation studies further validate the effectiveness of each proposed component. --- ### Instructions Shape Production of Language, not Processing Authors: Andreas Waldis, Leshem Choshen, Yufang Hou, Yotam Perlitz Published: 2026-05-11 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.11206v2 Instructions trigger a production-centered mechanism in language models. Through a cognitively inspired lens that separates language processing and production, we reveal this mechanism as an asymmetry between the two stages by probing task-specific information layer-wise across five binary judgment tasks. Specifically, we measure how instruction tokens shape information both when sample tokens, the input under evaluation, are processed and when output tokens are produced. Across prompting variations, task-specific information in sample tokens remains largely stable and correlates only weakly with behavior, whereas the same information in output tokens varies substantially and correlates strongly with behavior. Attention-based interventions confirm this pattern causally: blocking instruction flow to all subsequent tokens reduces both behavior and information in output tokens, whereas blocking it only to sample tokens has minimal effect on either. The asymmetry generalizes across model families and tasks, and becomes sharper with model scale and instruction-tuning, both of which disproportionately affect the production stage. Our findings suggest that understanding model capabilities requires jointly assessing internals and behavior, while decomposing the internal perspective by token position to distinguish the processing of input tokens from the production of output tokens. --- ### The Scaling Law of Evaluation Failure: Why Simple Averaging Collapses Under Data Sparsity and Item Difficulty Gaps, and How Item Response Theory Recovers Ground Truth Across Domains Authors: Jung Min Kang Published: 2026-05-11 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.11205v1 Benchmark evaluation across AI and safety-critical domains overwhelmingly relies on simple averaging. We demonstrate that this practice produces substantially misleading rankings when two conditions co-occur: (1) the evaluation matrix is sparse and (2) items vary substantially in difficulty. Through controlled simulation experiments across four domains -- NLP (GLUE), clinical drug trials, autonomous vehicle safety, and cybersecurity -- we show that Spearman rank correlation $ρ$ between simple-average rankings and ground-truth rankings degrades from $ρ= 1.000$ at 100% coverage to $ρ= 0.809$ at 67% coverage with high difficulty heterogeneity (mean over 20 seeds). A standard two-parameter logistic (2PL) Item Response Theory (IRT) model maintains $ρ\geq 0.996$ across all conditions. A 150-condition grid sweep over sparsity $S \in (0, 0.70)$ and difficulty gap $D \in (0.5, 5.0)$ confirms that ranking error forms a failure surface with a strong $S \times D$ interaction ($γ_3 = +0.20$, $t = 13.05$), while IRT maintains $ρ\geq 0.993$ throughout. We discuss implications for Physical AI benchmarking, where evaluation matrices are often incomplete and difficulty gaps are extreme. --- ### Multi-Agent System Identification with Nonlinear Sheaf Diffusion Authors: Nivar Anwer, Hans Riess, Matthew Hale Published: 2026-05-11 Categories: eess.SY, cs.LG, cs.MA, math.AT arXiv: https://arxiv.org/abs/2605.11204v1 Local interaction laws governing multi-agent systems can be difficult to recover from trajectory data, even when the dynamics are observed faithfully. In systems governed by a nonlinear sheaf Laplacian -- a generalization of the graph Laplacian accommodating heterogeneous state spaces and asymmetric communication channels -- the coordination law is encoded by edge potential functions whose gradients produce the inter-agent forces. Because trajectory observations record node-state evolution, they expose only the aggregate effect of the edge forces at each node: distinct interaction laws that agree at the node level are indistinguishable from trajectory data alone. We show that the fundamental obstruction to recovery is topological, measured by sheaf cohomology, and that unique recovery from an unconstrained function class is possible if and only if this cohomology vanishes. When the obstruction is nontrivial, we show that recovery within a finite-dimensional parameterized class is possible precisely when a data-dependent information matrix is positive definite. Experiments validate the theory and illustrate that accurate trajectory reproduction need not certify recovery of the underlying interaction law. --- ### FeatMap: Understanding image manipulation in the feature space and its implications for feature space geometry Authors: Elias B. Krey, Nils Neukirch, Nils Strodthoff Published: 2026-05-11 Categories: cs.LG, cs.CV arXiv: https://arxiv.org/abs/2605.11203v1 Intermediate feature representations represent the backbone for the expressivity and adaptability of deep neural networks. However, their geometric structure remains poorly understood. In this submission, we provide indirect insights into this matter by applying a broad selection of manipulations in input space, ranging from geometric and photometric transformations to local masking and semantic manipulations using generative image editing models, and assess the feasibility of learning a mapping in the feature space, mapping from the original to the manipulated feature map. To this end, we devise different types of mappings, from linear to non-linear and local to global mappings and assess both the reconstruction quality of the mapping as well as the semantic content of the mapped representations. We demonstrate the feasibility of learning such mappings for all considered transformations. While global (transformer) models that operate on the full feature map often achieve best results, we show that the same can be achieved with a shared linear model operating on a single feature vector typically with very little degradation in reconstruction quality, even for highly non-trivial semantic manipulations. We analyze the corresponding mappings across different feature layers and characterize them according to dominance of weight vs. bias and the effective rank of the linear transformations. These results provide hints for the hypothesis that the feature space is to a first degree of approximation organized in linear structures. From a broader perspective, the study demonstrates that generative image editing models might open the door to a deeper understanding of the feature space through input manipulation. --- ### Continuous Discovery of Vulnerabilities in LLM Serving Systems with Fuzzing Authors: Yunze Zhao, Yibo Zhao, Yuchen Zhang, Zaoxing Liu, Michelle L. Mazurek Published: 2026-05-11 Categories: cs.CR, cs.AI, cs.LG, cs.SE arXiv: https://arxiv.org/abs/2605.11202v1 LLM inference and serving systems have become security-critical infrastructure; however, many of their most concerning failures arise from the serving layer rather than from model behavior alone. Modern inference engines combine KV cache, batching, prefix sharing, speculative decoding, adapters, and multi-tenant scheduling, creating shared-state behavior that only emerges under realistic concurrent workloads and is missed by standard model, safety, and API tests. We present GRIEF, a greybox fuzzer for LLM inference engines that treats timed multi-request traces as first-class inputs, uses lightweight oracles to detect crashes, hangs, performance pathologies, and silent output corruption, and applies controlled replay with log-probability checks to confirm reproducible serving-layer failures. Across early campaigns on vLLM and SGLang, GRIEF discovers 15 vulnerabilities, 10 confirmed by engine developers, including 2 CVEs, spanning KV-cache isolation failures, cross-request performance interference, and crash or liveness bugs. These results show that concurrency, caching, and state reuse can induce silent cross-request contamination, noisy-neighbor denial of service, and delayed crashes without malformed inputs or explicit server errors, making concurrent serving behavior a first-class security and reliability boundary for LLM infrastructure. --- ### On the Impact of Crossover in Many-Objective Optimization: A Runtime Analysis of NSGA-III Authors: Andre Opris Published: 2026-05-11 Categories: cs.NE arXiv: https://arxiv.org/abs/2605.11201v1 In recent years, a theoretical understanding has rapidly advanced regarding how popular multi-objective evolutionary algorithms (MOEAs) can optimize many-objective problems. However, the benefits of using crossover in many-objective optimization are theoretically not understood, except for specifically designed benchmark functions tuned to particular crossover operators, and still lag significantly behind its practical use. In this paper, we build upon this line of research and present a theoretical runtime analysis of the widely used NSGA-III algorithm on the classical $m$-objective $m$-OneJumpZeroJump function ($m$-OJZJ for short). Our results demonstrate that NSGA-III with crossover optimizes $m$-OJZJ asymptotically faster than NSGA-III without crossover for any number $m$ of objectives for huge parameter regimes. We complement our analysis by providing a lower runtime bound on $4$-OJZJ when crossover is turned off. --- ### Operator-Guided Model Reduction for Generative Sampling in Lattice Field Theory Authors: Moxian Qian Published: 2026-05-11 Categories: hep-lat, cs.LG arXiv: https://arxiv.org/abs/2605.11199v2 Neural generative samplers for lattice field theory can be costly to train and evaluate. When they miss modes or assign them incorrect relative weights, biased observables do not reveal which collective variables are responsible. We project a trained flow-matching velocity onto vector fields built from lattice operators and Fourier modes. In two-dimensional lattice $φ^4$ theory, the projection separates changes in the overall magnetization from the lowest nonzero-momentum fluctuations and guides an explicit invertible proposal that treats them separately. Allowing the amplitude of the lowest nonzero-momentum fluctuations to depend on the magnetization improves the overlap between the proposal and target distributions, while the same two-parameter modification at higher momenta gives smaller improvements. The Metropolis--Hastings correction defines a Markov chain with the target Boltzmann distribution as its stationary law, and the normalized proposal density yields finite-volume partition-function estimates consistent with an independent HMC calculation. At the larger tested volume, the overlap between the proposal and target distributions deteriorates substantially, limiting the range over which the same parameterization remains effective. --- ### Operator Spectroscopy of Trained Lattice Samplers Authors: Moxian Qian Published: 2026-05-11 Categories: hep-lat, cs.LG arXiv: https://arxiv.org/abs/2605.11199v1 Trained lattice samplers are usually judged by the ensembles they generate. Here we instead analyze the trained field-space function itself: a flow-matching velocity, a diffusion score, or a normalizing-flow action residual. We project these functions onto operator bases fixed before the fit, chosen from symmetry, exact Gaussian path limits, finite-volume modes, and gauge covariance. For two-dimensional lattice \(φ^4\), a trained straight-flow teacher is not described by a local force basis alone. After the local transport basis, the residual separates into a zero-mode Binder component and a lowest-shell finite-\(k\) correlator component. The deflated zero-mode polynomial \(P_5(M;t)\) reduces the dominant Binder-tail component, while \(φ^\perp_{|n|^2=1}\) reduces the finite-\(k\) correlator component; wrong-parity, off-zero-mode, and random controls do not produce the same reductions. The same projection distinguishes other sampler classes. Diffusion follows the force-resolvent ordering predicted by the free theory, reverse-KL normalizing-flow collapse appears as a forbidden odd zero-mode residual, and gauge-equivariant teachers are resolved by Wilson-loop-force tangent directions. The operator basis is model- and symmetry-dependent, but the test is common: project the trained field-space function and retain sectors that lower held-out residuals and pass the available controls. --- ### Variational Linear Attention: Stable Associative Memory for Long-Context Transformers Authors: Vishal Pandey, Gopal Singh Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11196v1 Linear attention reduces the quadratic cost of softmax attention to $\mathcal{O}(T)$, but its memory state grows as $\mathcal{O}(T)$ in Frobenius norm, causing progressive interference between stored associations. We introduce \textbf{Variational Linear Attention} (VLA), which reframes the memory update as an online regularised least-squares problem with an adaptive penalty matrix maintained via the Sherman-Morrison rank-1 formula. We prove that normalising the write direction to unit length gives the recurrence Jacobian spectral norm exactly $1$ for all sequence lengths and head dimensions (Proposition 2), and that the state norm is self-limiting under bounded inputs (Proposition 1). Empirically, VLA reduces $\|S_t\|_F$ by $109\times$ relative to standard linear attention at $T{=}1{,}000$, achieves near-perfect exact-match accuracy on multi-query associative recall within the effective per-head memory regime ($n_\text{pairs} < d_h$), maintaining substantially higher retrieval performance than DeltaNet and standard linear attention under increasing memory load, and maintains 62\% accuracy at the per-head capacity boundary. A Triton-fused kernel achieves $14\times$ speedup over sequential Python and $\mathcal{O}(T)$ scaling, crossing below softmax attention latency at approximately 43\,000 tokens. --- ### How Does Differential Privacy Affect Social Bias in LLMs? A Systematic Evaluation Authors: Eduardo Tenorio, Karuna Bhaila, Xintao Wu Published: 2026-05-11 Categories: cs.CL arXiv: https://arxiv.org/abs/2605.11195v1 Large language models (LLMs) trained on web-scale corpora can memorize sensitive training data, posing significant privacy risks. Differential privacy (DP) has emerged as a principled framework that limits the influence of individual data points during training, yet the relationship between differential privacy and social bias in LLMs remains poorly understood. To investigate this, we present a systematic evaluation of social bias in a pretrained LLM trained with DP-SGD, comparing a DP model against non-DP baselines across four complementary paradigms: sentence scoring, text completion, tabular classification, and question answering. We find that DP reduces bias in sentence scoring tasks, where bias is measured through controlled likelihood comparisons, yet this improvement does not generalize across all tasks. Our results reveal a discrepancy between logit-level bias and output-level bias. Moreover, decreasing memorization does not necessarily reduce unfairness, underscoring the importance of multi-paradigm evaluation when assessing fairness in LLMs. --- ### Exploring Token-Space Manipulation in Latent Audio Tokenizers Authors: Francesco Paissan, Luca Della Libera, Mirco Ravanelli, Cem Subakan Published: 2026-05-11 Categories: cs.SD, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.11192v1 Neural audio codecs provide compact discrete representations for speech generation and manipulation. However, most codecs organize tokens as frame-level sequences, making it difficult to study or intervene on global factors of variation. In this work, we propose the Latent Audio Tokenizer for Token-space Editing (LATTE) that appends a fixed set of learnable latent tokens to the audio feature sequence and retains only these tokens for quantization and decoding. This design produces a compact, non-temporally aligned bottleneck in which each token can aggregate global information across the full utterance. We show that the resulting tokenizer preserves competitive reconstruction quality in low-bitrate speech coding settings while enabling simple token-space interventions. In particular, we find that swapping selected latent token positions between utterances can modify global attributes, such as speaker identity and background noise, and we evaluate these interventions on voice conversion and denoising tasks. Our results suggest that compact latent audio tokenizers can support controllable audio manipulation without supervision in task-specific editing models. --- ### Adaptive Policy Learning Under Unknown Network Interference Authors: Aidan Gleich, Eric Laber, Alexander Volfovsky Published: 2026-05-11 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2605.11191v1 Adaptive experimentation under unknown network interference requires solving two coupled problems: (i) learning the underlying dynamics of interference among units and (ii) using these dynamics to inform treatment allocation in order to maximize a cumulative outcome of interest (e.g. revenue). Existing adaptive experimentation methods either assume the interference network is fully known or bypass the network by operating on coarse cluster-level randomizations. We develop a Thompson sampling algorithm that jointly learns the interference network and adaptively optimizes individual-level treatment allocations via a Gibbs sampler. The algorithm returns both an optimized treatment policy and an estimate of the interference network; the latter supports downstream causal analyses such as estimation of direct, indirect, and total treatment effects. For additive spillover models, we show that total reward is linear in the treatment vector with coefficients given by an $n$-dimensional latent score. We prove a Bayesian regret bound of order $\sqrt{nT \cdot B \log(en/B)}$ for exact posterior sampling; empirically, our Gibbs-based approximate sampler achieves regret consistent with this rate and remains sublinear when the additive spillovers assumption is violated. For general Neighborhood Interference, where this reduction is unavailable, we analyze an explore-then-commit variant with $O(n^2 \log T)$ graph-discovery cost. An information-theoretic $Ω(n \log T)$ lower bound complements both results. Empirically, our method achieves more than an order-of-magnitude reduction in regret in head-to-head comparisons. On two real-world networks, the algorithm achieves sublinear regret and yields downstream effect estimates with small RMSE relative to the truth. --- ### Deep Learning for Protein Complex Prediction and Design Authors: Ziwei Xie Published: 2026-05-11 Categories: cs.LG, q-bio.BM arXiv: https://arxiv.org/abs/2605.11189v1 Accurately modeling and designing protein complex structures is a central problem in computational structural biology, with broad implications for understanding cellular function and developing therapeutics. This thesis investigates two fundamental aspects of this problem using deep learning: domain-specific architectures that capture the hierarchical nature of protein structures, and search algorithms that efficiently navigate the vast sequence spaces of protein complexes to identify interacting homologs for improving complex structure prediction and to design protein sequences. --- ### Adversarial SQL Injection Generation with LLM-Based Architectures Authors: Ali Karakoc, H. Birkan Yilmaz Published: 2026-05-11 Categories: cs.CR, cs.AI, cs.ET arXiv: https://arxiv.org/abs/2605.11188v1 SQL injection (SQLi) attacks are still one of the serious attacks ranked in the Open Worldwide Application Security Project (OWASP) Top 10 threats. Today, with advances in Artificial Intelligence (AI), especially in Large Language Models (LLMs), an opportunity has been created for automating adversarial attack tests to measure the defense mechanisms. In this paper, we aim to create a comprehensive evaluation of use cases that utilize LLMs for adversarial SQL injection generation. We introduce two novel LLM-based systems, Retrieval Augmented Generation for Adversarial SQLi (RADAGAS) and Reflective Chain-of-Thought SQLi (RefleXQLi), and compare them with existing baselines against 10 Web Application Firewalls (WAFs) and one execution-based MySQL validator. To perform a comprehensive test, we used six rule-based open-source WAFs (ModSecurity PL1--3, Coraza PL1--3), 2 AI/ML-based WAFs (WAF Brain, CNN-WAF), and 2 commercial WAFs (AWS WAF and Cloudflare WAF). For the LLM models, we used GPT-4o, Claude 3.7 Sonnet, and DeepSeek R1. Our tests consist of 240 experiments that generate 240,000 payloads and perform 2.2 million tests against WAFs. Our comprehensive evaluation reveals that RADAGAS-GPT4o outperforms other baseline models with a 22.73\% bypass rate. The proposed RADAGAS variants are highly successful on AI/ML-based WAFs (92.49\% on WAF-Brain by RADAGAS-DeepSeek, 80.48\% on CNN-WAF by RADAGAS-Claude), but struggle to bypass rule-based WAFs (0--5.70\% on ModSecurity and Coraza). In addition to these findings, another observation is that creating less diverse payloads achieves more bypasses, however they show poor results if the initially chosen payload is not successful. We observe that our findings provide a comprehensive view on using LLM-based approaches in security testing. --- ### CATS: Cascaded Adaptive Tree Speculation for Memory-Limited LLM Inference Acceleration Authors: Yuning Han, Yangchenchen Jin, Dylan Zhao, Jingwei Sun Published: 2026-05-11 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.11186v1 Auto-regressive decoding in Large Language Models (LLMs) is inherently memory-bound: every generation step requires loading the model weights and intermediate results from memory (e.g., High-Bandwidth Memory (HBM) for GPU servers), making throughput bottlenecked by memory bandwidth rather than compute. Speculative decoding addresses this by enabling parallel verification of multiple draft tokens, effectively amortizing the cost of each target-model call. However, existing speculative decoding methods are designed under the assumption that HBM is sufficiently large to hold both the target model and an auxiliary draft model simultaneously -- an assumption that breaks down on memory-constrained devices such as edge platforms with limited DRAM. We analyze the inference bottleneck in this memory-limited regime and propose CATS, a self-speculative decoding framework that conducts cascaded verification and correction based on the memory budget and parameter offloading patterns on memory-limited devices. This design maximizes token acceptance rate and end-to-end speedup while keeping the peak memory footprint on the device equal to that of the target model alone. We evaluate CATS on different models across five benchmarks on real edge devices. CATS can achieve a wall-clock speedup of up to 5.08x with no degradation in generation quality, outperforming the SOTA method by up to 1.45x under edge memory constraints. --- ### The Many Faces of On-Policy Distillation: Pitfalls, Mechanisms, and Fixes Authors: Siqi Zhu, Xuyan Ye, Hongyu Lu, Weiye Shi, Ge Liu Published: 2026-05-11 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.11182v2 On-policy distillation (OPD) and on-policy self-distillation (OPSD) have emerged as promising post-training methods for large language models, offering dense token-level supervision on trajectories sampled from the model's own policy. However, existing results on their effectiveness remain mixed: while OP(S)D has shown promise in system prompt and knowledge internalization, recent studies also report instability and degradation. In this work, we present a comprehensive empirical study of when OPD and OPSD work, when they fail, and why. We find that OPD on mathematical reasoning is highly sensitive to teacher choice and loss formulation, whereas OPSD fails in our tested settings due to test-time absence of instance-specific privileged information (PI). In contrast, OPSD is effective when PI represents a shared latent rule, such as a system prompt or alignment preference. We identify three failure mechanisms: (1) distribution mismatch between teacher and student caused by conditioning on student-generated prefixes, (2) optimization instability from biased TopK reverse-KL gradients, and (3) an OPSD-specific limitation where the student learns a PI-free policy that aggregates PI-conditioned teachers, which is insufficient when PI is instance-specific. We further show that stop-gradient TopK objectives, RLVR-adapted teachers, and SFT-stabilized students mitigate these failures. --- ### Muon is Not That Special: Random or Inverted Spectra Work Just as Well Authors: Zakhar Shumaylov, Nathaël Da Costa, Peter Zaika, Bálint Mucsányi, Alex Massucco, Yoav Gelberg, Carola-Bibiane Schönlieb, Yarin Gal, Philipp Hennig Published: 2026-05-11 Categories: cs.LG, cs.AI, math.NA, math.OC, stat.ML arXiv: https://arxiv.org/abs/2605.11181v1 The recent empirical success of the Muon optimizer has renewed interest in non-Euclidean optimization, typically justified by similarities with second-order methods, and linear minimization oracle (LMO) theory. In this paper, we challenge this geometric narrative through three contributions, demonstrating that precise geometric structure is not the key factor affecting optimization performance. First, we introduce Freon, a family of optimizers based on Schatten (quasi-)norms, powered by a novel, provably optimal QDWH-based iterative approximation. Freon naturally interpolates between SGD and Muon, while smoothly extrapolating into the quasi-norm regime. Empirically, the best-performing Schatten parameters for GPT-2 lie strictly within the quasi-norm regime, and thus cannot be represented by any unitarily invariant LMO. Second, noting that Freon performs well across a wide range of exponents, we introduce Kaon, an absurd optimizer that replaces singular values with random noise. Despite lacking any coherent geometric structure, Kaon matches Muon's performance and retains classical convergence guarantees, proving that strict adherence to a precise geometry is practically irrelevant. Third, having shown that geometry is not the primary driver of performance, we demonstrate it is instead controlled by two local quantities: alignment and descent potential. Ultimately, each optimizer must tune its step size around these two quantities. While their dynamics are difficult to predict a-priori, evaluating them within a stochastic random feature model yields a precise insight: Muon succeeds not by tracking an ideal global geometry, but by guaranteeing step-size optimality. --- ### Interpretable Machine Learning for Spatial Science: A Lie-Algebraic Kernel for Rotationally Anisotropic Gaussian Processes Authors: Kane Warrior, Dalia Chakrabarty Published: 2026-05-11 Categories: stat.ML, cs.LG arXiv: https://arxiv.org/abs/2605.11179v1 Many three-dimensional spatial fields are anisotropic, with directions of rapid and slow variation that need not align with the coordinate axes. Standard Gaussian process kernels with Automatic Relevance Determination (ARD) capture only axis-aligned anisotropy, while generic full symmetric positive definite (SPD) metrics can represent rotated anisotropy but do not parameterise principal length-scales and directions directly. We introduce an interpretable rotationally anisotropic GP kernel that parameterises a three-dimensional SPD covariance metric using three principal length-scales and an explicit SO(3) rotation. The rotation is represented by an axis-angle vector and mapped to SO(3) via the Lie-algebra exponential map, giving unconstrained Euclidean coordinates for inference while always inducing a valid SPD metric. The construction spans the same family of three-dimensional SPD covariance metrics as a generic full-SPD parameterisation, but exposes the geometry differently: length-scales and orientation are explicit, interpretable, and directly available for prior specification and posterior summaries. We perform Bayesian inference on these quantities using Markov Chain Monte Carlo (MCMC), and characterise the resulting symmetries and weakly identified regimes. On synthetic data with rotated anisotropy, the posterior recovers the generating metric and improves prediction relative to an axis-aligned ARD baseline, while matching the predictive performance of a generic full SPD baseline. When the ground truth is axis-aligned, posterior mass concentrates near the identity rotation and predictive performance matches ARD. On a material-density dataset from a laboratory-fabricated nano-brick, the inferred metric reveals rotated anisotropy that is not captured by axis-aligned kernels. --- ### Oversmoothing as Representation Degeneracy in Neural Sheaf Diffusion Authors: Arif Dönmez, Axel Mosig, Ellen Fritsche, Katharina Koch Published: 2026-05-11 Categories: cs.LG, cs.AI, math.RT arXiv: https://arxiv.org/abs/2605.11178v1 Neural Sheaf Diffusion (NSD) generalizes diffusion-based Graph Neural Networks by replacing scalar graph Laplacians with sheaf Laplacians whose learned restriction maps define a task-adapted geometry. While the diffusion limit of NSD is known to be the space of global sections, the representation-theoretic structure of this harmonic space remains largely implicit. We develop a quiver-theoretic interpretation of NSD by identifying cellular sheaves on graphs with representations of the associated incidence quiver. Under this correspondence, learned sheaf geometries become points in a finite-dimensional representation space. We show that direct-sum decompositions of the underlying incidence-quiver representation induce decompositions of the harmonic space reached in the diffusion limit. This gives an algebraic interpretation of oversmoothing as representation degeneration: learned sheaves may collapse toward low-complexity summands whose global sections fail to preserve discriminative information. Building on this viewpoint, we connect sheaf diffusion to stability and moment-map principles from Geometric Invariant Theory. We introduce moment-map-inspired regularizers that bias restriction maps toward balanced representation geometries, and identify a structural obstruction in equal-stalk architectures: when $d_v = d_e$, admissibility for learnable stability parameters forces the trivial all-object summand onto a stability wall. Non-uniform stalk dimensions remove this obstruction, making adaptive stability meaningful. Experiments on heterophilic benchmarks are consistent with this mechanism: breaking stalk symmetry can reduce variance or improve validation behavior, and adaptive stability becomes more effective in selected rectangular settings. Overall, our framework reframes oversmoothing as a degeneration phenomenon in the representation geometry underlying learned sheaf diffusion. --- ### From raw data to neutrino candidates: a neural-network pipeline for Baikal-GVD Authors: A. Matseiko, G. Plotnikov, I. Kharuk Published: 2026-05-11 Categories: astro-ph.IM, cs.LG, hep-ex arXiv: https://arxiv.org/abs/2605.11176v1 We present a neural-network-based data processing pipeline for Baikal-GVD, designed to improve event reconstruction quality and accelerate neutrino candidates selection. The pipeline comprises three stages: fast suppression of extensive air shower events, suppression of noise optical modules activations, and extraction of high confidence neutrino candidates. All three networks employ a transformer architecture that exploits inter-hit correlations through the attention mechanism. Applied sequentially, the pipeline achieves orders-of-magnitude speedup over the standard reconstruction chain. Moreover, noise suppression neural network surpasses the accuracy of algorithmic noise suppression algorithms and provides estimate for time residuals of the signal hits, which is crucial for identification of track-like hits. We address the domain shift between Monte Carlo simulations and experimental data by incorporating a domain adaptation technique, demonstrating improved agreement between the two domains. The resulting framework enables near-real-time event classification, with direct applications to multi-messenger alert systems and diffuse neutrino flux measurements. --- ### Optimistic Dual Averaging Unifies Modern Optimizers Authors: Thomas Pethick, Wanyun Xie, Roman Machacek, Volkan Cevher Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11172v1 We introduce SODA, a generalization of Optimistic Dual Averaging, which provides a common perspective on state-of-the-art optimizers like Muon, Lion, AdEMAMix and NAdam, showing that they can all be viewed as optimistic instances of this framework. Based on this framing, we propose a practical SODA wrapper for any base optimizer that eliminates weight decay tuning through a theoretically-grounded $1/k$ decay schedule. Empirical results across various scales and training horizons show that SODA consistently improves performance without any additional hyperparameter tuning. --- ### Unlearning with Asymmetric Sources: Improved Unlearning-Utility Trade-off with Public Data Authors: Ahmed Mehdi Inane, Vincent Quirion, Gintare Karolina Dziugaite, Ioannis Mitliagkas Published: 2026-05-11 Categories: cs.LG, cs.CR arXiv: https://arxiv.org/abs/2605.11170v2 Noise-based certified machine unlearning currently faces a hard ceiling: the noise magnitude required to certify unlearning typically destroys model utility, particularly for large-scale deletion requests. While leveraging public data is a standard technique in differential privacy to relax this tension, its role in unlearning remains unexplored. We address this gap by introducing Asymmetric Langevin Unlearning (ALU), a framework that uses public data to mitigate privacy costs. We prove that public data injection suppresses the unlearning cost by a factor of $O(1/n_{\mathrm{pub}}^2)$, guaranteeing a strict computational advantage over retraining. This establishes a new control mechanism: practitioners can mitigate the need for high noise-and the associated utility loss-by increasing the volume of public data. Crucially, we analyze the realistic setting of distribution mismatch, explicitly characterizing how shifts between public and private sources impact utility. We show that ALU enables mass unlearning of constant dataset fractions -- a regime where standard symmetric methods become impractical -- while maintaining high utility. Empirical evaluations using variational Rényi divergence and membership inference attacks confirm that ALU effectively thwarts privacy attacks while preserving utility under reasonable distribution shifts. --- ### OLIVIA: Online Learning via Inference-time Action Adaptation for Decision Making in LLM ReAct Agents Authors: Sheldon Yu, Junda Wu, Xintong Li, Nikki Lijing Kuang, Sizhe Zhou, Tong Yu, Jiawei Han, Jingbo Shang, Julian McAuley Published: 2026-05-11 Categories: cs.AI arXiv: https://arxiv.org/abs/2605.11169v1 Large language model agents interleave reasoning, action selection, and observation to solve sequential decision-making tasks. In deployed settings where agents repeatedly handle related multi-step tasks, small action-selection errors can accumulate into wasted tool calls, latency, and reduced reliability. Despite this need for deployment-time improvement, existing inference-time adaptation methods for LLM agents mainly rely on prompting or retrieval, which influence behavior indirectly through context manipulation. For ReAct-style agents, such approaches do not expose an explicit decision layer that can score candidate actions, represent uncertainty, or be updated online from action-level feedback. As a result, they provide limited support for trackable, fine-grained, and uncertainty-aware adaptation during deployment. We propose OLIVIA, an inference-time action adaptation framework for ReAct-style agents. OLIVIA models the LLM's final action-selection layer as a contextual linear bandit over candidate actions, with frozen hidden states as decision contexts. This choice is particularly suitable for deployment because it adapts behavior directly at the action-selection interface, preserves the underlying reasoning process, and provides explicit uncertainty estimates and lightweight online updates from action-level feedback. With upper-confidence-bound exploration, OLIVIA improves the policy sample-efficiently with minimal computational overhead. We instantiate OLIVIA on four benchmarks and show that it consistently improves task performance over static ReAct and prompt-based inference-time baselines. Our results suggest that explicit online decision layers provide an effective alternative to purely prompt- or retrieval-based adaptation for LLM agents during deployment. --- ### Variational predictive resampling Authors: Laura Battaglia, Stefano Cortinovis, Chris Holmes, David T. Frazier, Jack Jewson Published: 2026-05-11 Categories: stat.ME, stat.CO, stat.ML arXiv: https://arxiv.org/abs/2605.11168v2 Bayesian inference provides principled uncertainty quantification, but accurate posterior sampling with MCMC can be computationally prohibitive for modern applications. Variational inference (VI) offers a scalable alternative and often yields accurate predictive distributions, but cheap variational families such as mean-field (MF) can produce over-concentrated approximations that miss posterior dependence. We propose variational predictive resampling (VPR), a scalable posterior sampling method that exploits VI's predictive strength within a predictive-resampling framework to better approximate the Bayesian posterior. Given a prior-likelihood pair, VPR repeatedly imputes future observations from the current variational predictive, updates the variational approximation after each imputation, and records the parameter value implied by the completed sample. We establish conditions under which the law of the parameter returned by VPR is well defined and show that its finite-horizon approximation converges to this limit. In a tractable Gaussian location model, we show that VPR with MF variational predictives converges to the exact Bayesian posterior, whereas the optimal MF-VI approximation retains a non-vanishing asymptotic gap. Experiments on linear regression, logistic regression, and hierarchical linear mixed-effects models demonstrate that VPR substantially improves posterior uncertainty quantification and recovers posterior dependence missed by MF-VI, while remaining computationally competitive with, and often more efficient than, MCMC. --- ### The Bicameral Model: Bidirectional Hidden-State Coupling Between Parallel Language Models Authors: Cedric Flamant, Udaya Ghai, Kanna Shimizu Published: 2026-05-11 Categories: cs.CL, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.11167v1 Existing multi-model and tool-augmented systems communicate by generating text, serializing every exchange through the output vocabulary. Can two pretrained language models instead coordinate through a continuous, concurrent channel? The Bicameral Model couples two frozen language models through a trainable neural interface on their intermediate hidden states. At every generation step, both models run in lockstep: a primary model drives the task while an auxiliary model operates tools, solves constraints, or executes code, with both conditioning on each other's activations through a translation network and a learned suppression gate ($\sim$1\% of combined parameters). The gate learns a selective communication protocol from task loss alone, without a prescribed format. We demonstrate the mechanism across three tool backends. On arithmetic, coupling two 0.5B models with a calculator raises accuracy from 36\% to 96\%. On logic grid puzzles, coupling two 0.6B models with a Z3 solver achieves $1.7\times$ the unaugmented baseline on ZebraLogic. On mathematical reasoning, coupling with a Python sandbox enables the auxiliary to generate problem-specific code from hidden-state signals alone, without ever seeing the problem text. --- ### Unpacking the Eye of the Beholder: Social Location, Identity, and the Moving Target of Political Perspectives Authors: Elena Sirotkina Published: 2026-05-11 Categories: cs.CV arXiv: https://arxiv.org/abs/2605.11166v1 Political and social identities structure how people evaluate political information, a finding decades deep in political science and routinely discarded by computational tools that often produce single scores that treat a piece of text, an image, or a video as if it means the same thing to everyone. This paper shows that it does not, and that the difference is consequential. To address this problem, I develop the Perspectivist Visual Political Sentiment (PVPS) classifier, which learns from approximately 82,000 evaluations by 5,575 U.S. adults to predict how audiences defined by political and social identities will evaluate the same image. Unlike standard tools that average systematic disagreement away, PVPS preserves it, returning an evaluative profile that records who agrees, who diverges, and along which identity lines. Applied to several influential studies of visual sentiment, PVPS shows that perceived violence in protest imagery and the emotional mechanisms behind protest image engagement both change substantively once audience identity is taken into account. It follows that what a political image conveys is a moving target, and measuring it requires knowing whom it is moving. --- ### COSMOS: Model-Agnostic Personalized Federated Learning with Clustered Server Models and Pseudo-Label-Only Communication Authors: Ben Rachmut, Luise Ge, William Yeoh, Ning Zhang, Yevgeniy Vorobeychik Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11165v3 Federated learning (FL) in heterogeneous environments remains challenging because client models often differ in both architecture and data distribution. While recent approaches attempt to address this challenge through client clustering and knowledge distillation, simultaneously handling architectural and statistical heterogeneity remains difficult. We introduce COSMOS, a model-agnostic framework that enables server-side personalization using only pseudo-label communication. Clients train local models and predict on the public data; the server clusters clients by prediction similarity, trains a cluster-specific model for each group using its own compute, and distills the resulting models back to clients. We provide the first theoretical analysis showing that distillation from the learned cluster models can yield exponential personalization risk contraction, going beyond the convergence-to-stationarity guarantees typically provided in model-agnostic FL. Experiments across benchmarks demonstrate that COSMOS consistently outperforms all model-agnostic FL baselines while remaining competitive with state-of-the-art personalized FL methods. More broadly, our results highlight personalized server-side learning with pseudo-labels as a promising paradigm for scalable and model-agnostic federated learning in highly heterogeneous environments. --- ### Benchmarking LLM-Based Static Analysis for Secure Smart Contract Development: Reliability, Limitations, and Potential Hybrid Solutions Authors: Stefan-Claudiu Susan, Andrei Arusoaie, Dorel Lucanu Published: 2026-05-11 Categories: cs.CR, cs.AI arXiv: https://arxiv.org/abs/2605.11163v1 The irreversible nature of blockchain transactions makes the identification of smart contract vulnerabilities an essential requirement for secure system development. While Large Language Models (LLMs) are increasingly integrated into developer workflows, their reliability as autonomous security auditors remains unproven. We assess whether current generative models are a viable replacement for, or only a complement to, traditional static-analysis tools. Our findings indicate that LLM efficacy is undermined by both inherent lexical bias and a lack of rigorous validation of external data inputs. This reliance on non-semantic heuristics, such as identifier naming, leads to a high frequency of false positives. Furthermore, prompting techniques reveal a trade-off between precision and recall. These results were derived using our custom automated framework, which achieves 92% accuracy in classifying model outputs. --- ### Interpretability Can Be Actionable Authors: Hadas Orgad, Fazl Barez, Tal Haklay, Isabelle Lee, Marius Mosbach, Anja Reusch, Naomi Saphra, Byron Wallace, Sarah Wiegreffe, Eric Wong, Ian Tenney, Mor Geva Published: 2026-05-11 Categories: cs.LG, cs.AI arXiv: https://arxiv.org/abs/2605.11161v1 Interpretability aims to explain the behavior of deep neural networks. Despite rapid growth, there is mounting concern that much of this work has not translated into practical impact, raising questions about its relevance and utility. This position paper argues that the central missing ingredient is not new methods, but evaluation criteria: interpretability should be evaluated by actionability--the extent to which insights enable concrete decisions and interventions beyond interpretability research itself. We define actionable interpretability along two dimensions--concreteness and validation--and analyze the barriers currently preventing real-world impact. To address these barriers, we identify five domains where interpretability offers unique leverage and present a framework for actionable interpretability with evaluation criteria aligned with practical outcomes. Our goal is not to downplay exploratory research, but to establish actionability as a core objective of interpretability research. --- ### CORE: Cyclic Orthotope Relation Embedding for Knowledge Graph Completion Authors: Yingqi Zeng, Luying Wang, Huiling Zhu Published: 2026-05-11 Categories: cs.LG arXiv: https://arxiv.org/abs/2605.11159v1 Knowledge graph completion (KGC) aims to automatically infer missing facts in multi-relational data by mapping entities and relations into continuous representation spaces. Recent region-based embedding models have shown great promise in capturing complex logical patterns by representing relations as geometric regions. However, these models inevitably suffer from absolute boundary constraints during optimization. Conversely, without such constraints, relation regions expand indefinitely. To address the limitation, we propose \textbf{CORE} (Cyclic Orthotope Relation Embedding), a novel KGC model that embeds entities and relations onto a boundary-less torus manifold.CORE represents relations as cyclic orthotopes on the torus manifold, allowing regions to seamlessly wrap around spatial boundaries to ensure smooth gradient conduction. Furthermore, an adaptive width regularization is introduced to prevent unconditional region expansion. Theoretical analysis proves that CORE can capture various complex relation patterns such as subsumption and intersection. Extensive experiments on four benchmark datasets demonstrate that CORE achieves highly competitive performance, significantly improving link prediction accuracy in dense semantic environments. --- ### The Price of Proportional Representation in Temporal Voting Authors: Nicholas Teh Published: 2026-05-11 Categories: cs.GT, cs.AI, cs.LG, cs.MA, econ.TH arXiv: https://arxiv.org/abs/2605.11157v1 We study proportional representation in the temporal voting model, where collective decisions are made repeatedly over time over a fixed horizon. Prior work has extensively investigated how proportional representation axioms from multiwinner voting (e.g., justified representation (JR) and its variants) can be adapted, satisfied, and verified in this setting. However, much less is understood about their interaction with social welfare. In this work, we quantify the efficiency cost of enforcing proportionality. We formalize the welfare-proportionality tension via the worst-case ratio between the maximum achievable utilitarian welfare and the maximum welfare attainable subject to a proportionality axiom. We show that imposing proportional representation in the temporal setting can incur a growing, yet sublinear, welfare loss as the number of voters or rounds increases. We further identify a clean separation among axioms: for JR, the welfare loss diminishes as the time horizon grows and vanishes asymptotically, whereas for stronger axioms this conflict persists even with many rounds. Moreover, we prove that welfare maximization under each axiom is NP-complete and APX-hard, even under static preferences and bounded-degree approvals, and provide fixed-parameter algorithms under several natural structural parameters. --- ### Quantifying the Reconstructability of Astrophysical Methods with Large Language Models and Information Theory: A Case Study in Spectral Reconstruction Authors: Hsing Wen Lin, Zong-Fu Sie Published: 2026-05-11 Categories: astro-ph.IM, cs.AI, cs.LG arXiv: https://arxiv.org/abs/2605.11154v2 Modern astrophysical studies rely heavily on complex data analysis pipelines; however, published descriptions often lack the detail required for computational reproducibility. In this work, we present an information-theoretic framework to quantify how effectively a method can be reconstructed from its written description. By treating algorithmic reconstruction as a probability distribution generated by Large Language Models (LLMs), we utilize Shannon entropy and Jensen-Shannon divergence to measure how strongly text constrains the hypothesis space of valid implementations. We demonstrate this approach through a case study of Trans-Neptunian Object (TNO) spectral reconstruction from sparse photometry. By prompting frontier LLMs with varying levels of manuscript text (Title, Abstract, and Methods), we find that while increasing text successfully clarifies the overall algorithmic structure, it fails to eliminate variance at the implementation level. This persistent variance establishes an "entropy floor," demonstrating that multiple divergent implementations remain consistent with explicit instructions. To evaluate practical reproducibility, we convert these reconstructed algorithms into executable pipelines. Our results reveal that, while LLMs easily recover core functional methodologies, they systematically fail to infer the tacit expert knowledge required for strict scientific calibration. This pilot study demonstrates that LLMs can be repurposed as a zero-shot diagnostic tool to audit methodological transparency, helping authors identify missing structural constraints and preserve scientific integrity in an era of automated research. --- ### Decomposing Evolutionary Mixture-of-LoRA Architectures: The Routing Lever, the Lifecycle Penalty, and a Substrate-Conditional Boundary Authors: Ramchand Kumaresan Published: 2026-05-11 Categories: cs.CL, cs.LG, cs.NE arXiv: https://arxiv.org/abs/2605.11153v1 We decompose an evolutionary mixture-of-LoRA system on a from-scratch ~150M-parameter widened-D substrate (D=1536, V=32000; D/V approx 0.048; the "widened-1536" substrate) into three factors -- a router rewrite (parallel sigmoid gate with learnable per-adapter floor and bounded temperature anneal, fed post-stack hidden states rather than token-embedding means), a per-domain leave-one-out evaluation scope, and a lifecycle of death plus alpha-blend inheritance plus SVD mutation plus slot reallocation -- and report a 5-of-8 partial 2^3 factorial run at n=3 seeds and 25000 adaptation steps per cell. The attribution chain is sharp on this substrate: the router rewrite carries the entire +0.0426 nat balanced log-PPL improvement (Delta = log PPL_ref - log PPL_test, positive = improvement; t=12.86, p=0.006) attributed to "the full evolutionary system vs the static B3 baseline"; the headline full-system-vs-B3 balanced contrast itself is +0.015 nats, t=1.94, p=0.19 at n=3 and does not clear alpha=0.05. The per-domain evaluation scope is null at seed-resolution, and the lifecycle is a net drag of approx -0.028 nats (t=-4.46,p=0.047 in the primary chain). An auxiliary alpha=0 inheritance counterfactual at n=3 seeds is sign-inconsistent at the headline metric and underpowered for either an equivalence or load-bearing conclusion (corrected from an earlier arithmetic-mean aggregator that erroneously cleared inheritance; see Appendix B.11). A base-perturbation probe directionally refutes a "genomic-context" reframe of the lifecycle role. A controllable synthetic sandbox locates a substrate-conditional regime boundary: evolutionary search on the routing channel is load-bearing only when adapters are pre-aligned to the task; in every other regime tested it underperforms, ties, or actively degrades the gradient solution. --- ### RankQ: Offline-to-Online Reinforcement Learning via Self-Supervised Action Ranking Authors: Andrew Choi, Wei Xu Published: 2026-05-11 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2605.11151v3 Offline-to-online reinforcement learning (RL) improves sample efficiency by leveraging pre-collected datasets prior to online interaction. A key challenge, however, is learning an accurate critic in large state--action spaces with limited dataset coverage. To mitigate harmful updates from value overestimation, prior methods impose pessimism by down-weighting out-of-distribution (OOD) actions relative to dataset actions. While effective, this essentially acts as a behavior cloning anchor and can hinder downstream online policy improvement when dataset actions are suboptimal. We propose RankQ, an offline-to-online Q-learning objective that augments temporal-difference learning with a self-supervised multi-term ranking loss to enforce structured action ordering. By learning relative action preferences rather than uniformly penalizing unseen actions, RankQ shapes the Q-function such that action gradients are directed toward higher-quality behaviors. Across sparse-reward D4RL benchmarks, RankQ achieves competitive overall performance against seven baselines. In vision-based robot learning, RankQ enables effective offline-to-online fine-tuning of a pretrained vision-language-action (VLA) model in a low-data regime, achieving an average simulation success rate 38.2 percentage points higher than the next best method. In a high-data setting, RankQ improves simulation performance by 13.7 percentage points over the next best method and demonstrates strong sim-to-real transfer, increasing real-world cube stacking success from 43.1% to 88.9% relative to the VLA's initial performance. --- ### RankQ: Offline-to-Online Reinforcement Learning via Self-Supervised Action Ranking Authors: Andrew Choi, Wei Xu Published: 2026-05-11 Categories: cs.AI, cs.RO arXiv: https://arxiv.org/abs/2605.11151v2 Offline-to-online reinforcement learning (RL) improves sample efficiency by leveraging pre-collected datasets prior to online interaction. A key challenge, however, is learning an accurate critic in large state--action spaces with limited dataset coverage. To mitigate harmful updates from value overestimation, prior methods impose pessimism by down-weighting out-of-distribution (OOD) actions relative to dataset actions. While effective, this essentially acts as a behavior cloning anchor and can hinder downstream online policy improvement when dataset actions are suboptimal. We propose RankQ, an offline-to-online Q-learning objective that augments temporal-difference learning with a self-supervised multi-term ranking loss to enforce structured action ordering. By learning relative action preferences rather than uniformly penalizing unseen actions, RankQ shapes the Q-function such that action gradients are directed toward higher-quality behaviors. Across sparse reward D4RL benchmarks, RankQ achieves performance competitive with or superior to seven prior methods. In vision-based robot learning, RankQ enables effective offline-to-online fine-tuning of a pretrained vision-language-action (VLA) model in a low-data regime, achieving on average a 42.7% higher simulation success rate than the next best method. In a high-data setting, RankQ improves simulation performance by 13.7% over the next best method and achieves strong sim-to-real transfer, increasing real-world cube stacking success from 43.1% to 88.9% relative to the VLA's initial performance. --- ### Debiasing Message Passing to Mitigate Popularity Bias in GNN-based Collaborative Filtering Authors: Md Aminul Islam, Ahmed Sayeed Faruk, Sourav Medya, Elena Zheleva Published: 2026-05-11 Categories: cs.IR, cs.LG arXiv: https://arxiv.org/abs/2605.11145v1 Collaborative filtering (CF) models based on graph neural networks (GNNs) achieve strong performance in recommender systems by propagating user-item signals over interaction graphs. However, they are highly susceptible to popularity bias, since skewed interaction distributions and repeated message passing across high-order neighborhoods amplify the influence of popular items while suppressing long-tail ones. Existing debiasing approaches, including re-weighting objectives, regularization, causal methods, and post-processing, are less effective in GNN-based settings because they do not directly counteract bias propagated through the aggregation process, and recent in-aggregation weighting methods often rely on static heuristics or unstable embedding estimates. We propose Debiasing Popularity Amplification in Aggregation (DPAA), a popularity debiasing framework for GNN-based CF that integrates adaptive, embedding-aware interaction weighting and layer-wise weighting directly into message passing. DPAA assigns interaction-level weights from a representation-aware popularity signal, stabilized by a smooth transition from pre-trained to evolving model embeddings during training. It further introduces a layer-wise weighting that amplifies higher-order neighborhoods, surfacing long-range interactions with diverse and underexposed items. Experiments on real-world and semi-synthetic datasets show that DPAA outperforms state-of-the-art popularity-bias correction methods for GNN-based CF. ---