Machine Learning
Not all solutions are created equal: An analytical dissociation of functional and representational similarity in deep linear neural networks
Lukas Braun, Erin Grant and Andrew M. Saxe
A foundational principle of connectionism is that perception, action, and cognition emerge from parallel computations among simple, interconnected units that generate and rely on neural representations. Accordingly, researchers employ multivariate pattern analysis to decode and compare the neural codes of artificial and biological networks, aiming to uncover their functions. However, there is limited analytical understanding of how a network's representation and function relate, despite this being essential to any quantitative notion of underlying function or functional similarity. We address this question using analysable two-layer linear networks and numerical simulations in non-linear networks. We find that function and representation are dissociated, allowing representational similarity without functional similarity and vice versa. Further, we show that neither robustness to input noise nor the level of generalization error constrain representations to the task. In contrast, networks robust to parameter noise have limited representational flexibility and must employ task-specific representations. Our findings suggest that representational
alignment reflects computational advantages beyond functional
alignment alone, with significant implications for interpreting and comparing the representations of connectionist systems.
Read the full abstract ▾
Computation and Language
Settle: Learning When to Stop Reasoning
Ryan Brown, Zihao Fu and Chris Russell
Reasoning models often continue generating after their answers have settled. Settle learns when to stop from answer stability in completed traces. It trains the existing end-of-
reasoning token while keeping other predictions close to the base model, and requires only ordinary decoding at inference. On MATH-500 with Qwen3-4B, Settle reduces
token count by 40% with a 0.5-percentage-point decrease in accuracy. It gains 6.16 percentage points over supervised fine-tuning on the same traces shortened at their first stable answer, at nearly identical
token counts. Its stopping score predicts whether a correct answer will remain correct. Settle extends the accuracy-
token-count Pareto frontier of the evaluated stopping methods.
Read the full abstract ▾
Computation and Language
When Clipping Reverses Correction: Failure Dynamics of Pointwise Forward-KL On-Policy Self-Distillation
Di Huang, Hao Li, Yixin Chen et al. (+1 more)
On-policy self-
distillation (OPSD) trains a student on its own generated responses using feedback from the same model conditioned on privileged information. On mathematical
reasoning, the original OPSD study finds that stylistic
tokens can dominate the training signal over math-related
tokens, and that pointwise clipping of the forward KL objective stabilizes training. Pointwise clipping caps each vocabulary-wise forward KL term at a fixed threshold before summing over the vocabulary. Follow-up studies have adopted this clipping, but its effect on training has not been directly examined. In matched training runs differing only in whether clipping is applied, we observe that clipped runs produce substantially more repetitions that persist to the end of the response than their unclipped counterparts. We trace this failure to the clipped objective. We prove that the clipped objective can fail to correct the student toward the teacher and can instead push clipped and unclipped
token probabilities away from its teacher. Our training runs agree with this analysis: inside repetitions, the clipped student places less probability than its teacher on leaving the repetition, and more on continuing it, whereas the unclipped runs stay close to their teachers.
Read the full abstract ▾
All authors Di Huang, Hao Li, Yixin Chen, Fuhai Li
Machine Learning
Anchoring Adversarial Trajectories to Data Manifolds: A Bilevel Transfer Optimization Framework
Yaohua Liu, Yifan Guo and Jiaxin Gao
A key bottleneck in
adversarial transfer is a trajectory-level
geometric disconnect: ambient gradients often drift away from the intrinsic data
manifold, causing surrogate-specific overfitting. To rectify this, we propose
Manifold Anchored Bilevel Transfer (MABT), a unified framework that anchors
adversarial trajectories to the shared semantic subspace. MABT introduces a relaxed
manifold-anchoring operator as a semantic rectifier to suppress off-
manifold noise. With this constraint, we cast transfer attack generation as a distributional bilevel optimization problem that learns a geometry-aligned initialization by minimizing expected transfer risk under a surrogate uncertainty distribution. We further develop a Hessian-free solver with linear-time complexity to handle the resulting hierarchy. Experiments demonstrate improved transferability for 10 baseline attackers across 28 attack configurations, diverse victim architectures, and defense mechanisms.
Read the full abstract ▾
Robotics
Cue the Flow: Steering Flow-Matching Policies for Open-World Delivery Manipulation
Haoxuan Wang, Griffin Galimi, Junhua Huang et al. (+4 more)
Open-world goods delivery requires mobile manipulators to follow free-form user instructions and manipulate potentially novel objects. Existing dual-system approaches use high-level grounding models to convert
language into grounded visual
prompts, but their low-level controllers can remain brittle under noisy perception, dynamic scenes, and contact-rich interactions. We instead use a pretrained flow-matching
vision-
language-action model as the low-level control interface, leveraging its reactivity and robustness to environmental changes while treating the grounding output as a spatial cue for policy steering. Our key insight is that the pretrained VLA already provides a strong manipulation prior, while the spatial cue supplies the missing target information needed to guide actions under novel
language–object mappings. Concretely, we introduce a lightweight cue-conditioned adapter. The adapter is first trained with contrastive objectives to produce salient and spatially discriminative cue representations, and is then supervised to predict a diagonal affine transformation over the generated action chunk, aligning policy steering with the cued target. Across tabletop and mobile-base settings, our method improves instruction following and manipulation success on both in-domain and out-of-domain objects, achieving up to near $2\times$ improvement in average task success rate with negligible inference overhead.
Read the full abstract ▾
All authors Haoxuan Wang, Griffin Galimi, Junhua Huang, Selina Song, Wayne Wu, Yan Yan, Bolei Zhou
Social and Information Networks · Cross-listed from math.DS
Community Detection using the Edge Laplacian Dynamics vis-a-vis Ricci Flow
Pratibha Bhandari and Soumyendu Raha
It has been found that utilizing the
geometric properties of the graph dynamics can bring out crucial information of the data that a statistical analysis will not. The properties of
Ricci flow bring out hidden dynamics of the data in the same way as decomposition of smooth
manifolds. In this paper we explore an alternative to
Ricci flow, where we find that we can bring out similar properties like it using a simpler and well defined Edge
Laplacian to practice community detection in graphs. Moreover, the results obtained by the present approach show that a flow driven by the Edge
Laplacian yields results similar to those of the
Ricci flow. In contrast, computing the Forman-
Ricci curvature requires looping over the adjacent edges, while computing Olivier-
Ricci curvature involves the computation of Wasserstein distance and probability distribution associated with the graph nodes, making the Edge
Laplacian approach computationally more efficient.
Read the full abstract ▾
Machine Learning
Smaller Models, Better Rejects: Preference Distillation Scaling
Rui Cai, Wenhui Zhu, Xiwen Chen et al. (+15 more)
Preference
distillation typically treats a teacher response as preferred and the student's own response as rejected. This assumes that self-generated failures are the most informative negatives and that rejects must come from a model at least as large as the student, making generation costly at scale. We find neither assumption holds: across students from 7B to 72B, smaller frozen models generate rejects with less inference compute yet train stronger students than self-generated rejects, before and after sequence-level knowledge
distillation, on code generation and mathematical
reasoning. To explain this result, we derive a finite-horizon utility bound for Direct Preference Optimization in a linearized feature model. The bound characterizes favorable reject distributions and motivates three interventions. First, mixing rejects from smaller and student-scale models improves performance as the smaller model's share increases. Second, reassigning rejects to other
prompts and shuffling their code
tokens still outperform length-matched gibberish, showing that task structure contributes to reject utility. Third, selecting candidates with lower likelihood under the reference policy improves net transfer when higher-likelihood candidates provide less useful contrast. Lower-likelihood selections outperform higher-likelihood ones for every source. These results suggest that effective rejects preserve task structure while limiting coupling to the reference policy, and that smaller frozen models can provide them at low cost.
Read the full abstract ▾
All authors Rui Cai, Wenhui Zhu, Xiwen Chen, Jincheng Cao, Han Yu, Shayan Mohajer Hamidi, Zelin He, Qiyao Ma, Daiwei Chen, Xuanzhao Dong, Yuanda Xu, Jelena Markovic-Voronov, Kayhan Behdin, Zhengze Zhou, Ran He, Alborz Geramifard, Rohit Jain, Zhe Zhao
Computer Vision
MeshOctave: Vertex Split-and-Rewire Cascades for Native Mesh Generation
Junkai Lin, Tianhao Zhao, Hang Long et al. (+10 more)
Generating compact, artist-style meshes with explicit
topology typically relies on autoregressive models which incur prohibitive sequential per-
token costs, or continuous flow models that depend on heuristic connectivity decoders. Next-scale generation paradigms offer a compelling alternative by enabling parallel intra-scale
token prediction and coarse-to-fine refinement from global structure to local
topology; yet, existing methods derive hierarchical scales via progressive mesh simplification and invert them sequentially. This eliminates intra-scale parallelism and scales generation steps linearly with face count. In this paper, we propose MeshOctave, which instead defines scale through dyadic spatial grid resolutions, framing coarsening as a deterministic collapse that merges vertices sharing a voxel
cell and inherits connectivity. Its inverse operation, split-and-rewire, determines which octant sub-vertices are instantiated for each coarse face and resolves local connectivity using discrete structural
tokens. These per-face operations require no serialization, each scale transition is modeled as an unordered set that adds one bit of coordinate precision, naturally supporting dynamic-length meshes and adaptive resolution refinement. We construct a scale-conditioned masked-uniform discrete
diffusion model to learn split-and-rewire operation from resolution collapse hierarchies. MeshOctave outperforms strong baselines in
geometric fidelity and topological validity by a non-trivial margin, while supporting adaptive resolution refinement and extending naturally to mesh subdivision tasks.
Read the full abstract ▾
All authors Junkai Lin, Tianhao Zhao, Hang Long, Huipeng Guo, Jielei Zhang, Youjia Zhang, Jiale Xu, Wenbing Li, Rendong Liang, Jozef Hladký, Matthias Nießner, Yuanming Hu, Wei Yang
Robotics
Sparse-WAM: Accelerating World Action Models via Action-Guided Sparse Imagination
Xinling Xie, Haodong Wang, Jiazhi Mi et al. (+8 more)
World-action models (WAMs) leverage pretrained
video models to improve generalization in
robot control by jointly predicting future visual states and actions. This capability comes at a substantial inference cost, as dense future-frame
tokens are repeatedly processed during denoising. Prior methods address this by
token pruning that prioritizes visual fidelity to reduce denoising costs in
video diffusion models. However, these methods do not use action relevance to determine which future-frame
tokens to retain during joint denoising in WAMs. In this paper, we propose Sparse-WAM, a training-free framework for action-guided sparse imagination that selectively processes future-frame
tokens to accelerate WAM inference. We observe substantial overlap in the spatial distribution of
attention from action
tokens to future-frame
tokens (action-to-future
attention) between consecutive denoising steps, despite continued updates to the future representations. Motivated by this, we develop Action-Guided
Token Selection to retain frame-specific action-relevant regions together with cross-frame context. However, a naive implementation can incur
attention-scoring and
token-packing overhead that offsets the computational savings from
pruning. We therefore introduce Pilot, an efficient engine that reduces sparse inference overhead through lightweight scoring and cross-step reuse of
token selections. On LIBERO with FastWAM-Joint and RoboLab-120 with Cosmos 3 Edge, Sparse-WAM achieves inference speedups of approximately $2.0\times$ and $1.8\times$, respectively, over dense eager inference on an NVIDIA RTX 4090, while largely preserving task performance.
Read the full abstract ▾
All authors Xinling Xie, Haodong Wang, Jiazhi Mi, Zhiming Liu, Zicong Hong, Xiaoyi Pang, Qianli Liu, Yangjia Hu, Ying Chen, Zhengyang Yan, Song Guo
Cryptography and Security
Approval Laundering: Systematizing Approval–Execution Binding Failures in AI Coding-Agent Harnesses
Yang Wang
Modern AI coding-
agent harnesses (Claude Code, Codex CLI, Cursor) rest their security boundary on a largely unexamined assumption: that the action A a human approves is the same action A' the harness executes, where A is fixed by a stated policy for what a scope grant or session-scoped approval authorizes. We show this assumption fails systematically and reproducibly. We introduce Approval Laundering, a taxonomy of six failure modes by which a harness's enforcement mechanism silently substitutes A' for A after approval: Scope, Argument, Temporal, Tool, Delegation, and Semantic laundering. Unlike prior work that evaluates risk classifiers against static corpora or infers implicit authorization boundaries, we study credential-binding integrity: given an already-approved action, does the harness dispatch exactly that action? Instrumenting Claude Code's pre-execution mediation point (PreToolUse), we conduct a controlled, headless, repeated-measures study of all six classes (N=19-20 runs each), reporting a Bound-Gap Rate (BGR) with Wilson confidence intervals and inter-rater agreement (kappa=1.0). We prototype Approval
Token, a keyed capability Hk(principal,
agent_id, session_id, tool, arguments, scope, expiry) issued by a mediator that never returns the key to the
agent, evaluated via paired before/after replay of 118 runs (McNemar's exact test). The
token fully eliminates Delegation laundering and, for our seeded session-identity-mismatch construction, Temporal laundering (p<10^-5), but by design leaves Scope laundering unaffected and shows no significant reduction in Argument laundering (p=1): an honest negative result, since these two classes leave every recorded dispatch field unchanged, diverging one process level below what a field-only verifier can observe. We discuss implications for defenses that bind only at the tool-invocation boundary.
Read the full abstract ▾
Robotics
SimEX: Simulation-Integrated Robotics AutoResearch
Jiaheng Hu, Roberto Martin-Martin, Peter Stone et al. (+3 more)
Coding
agents powered by large
language models (LLMs) have shown remarkable abilities to autonomously reason about and achieve goals in the digital world. However, bringing this success to the physical world remains challenging. On the one hand, direct generation methods (e.g., Code as Policies) often suffer from the LLMs' insufficient understanding of
robots and physical environments. On the other hand, iterative trial-and-error tuning in the physical world (e.g., physical autoresearch) induces significant experimental cost and safety concerns. We introduce SimEX: Simulation-Integrated Robotics AutoResearch, an autoresearch framework that tightly integrates simulated experimentation, enabling coding
agents to efficiently acquire physical capabilities for controlling real
robots. SimEX operates in two stages. First, the
agent conducts open-ended probe-and-optimize iterations in simulation, developing a
robot toolbox with robust and generalizable capabilities. Second, the
agent adapts the toolbox and the simulator together through only a few physical trials: each trial corrects the simulator, and the corrected simulator is used to diagnose failures and screen candidate repairs. We evaluate SimEX extensively in sim-to-sim settings and on physical
robots. On challenging real-world manipulation tasks including towel folding, barcode scanning, and plate manipulation, SimEX enables coding
agents to efficiently acquire
robot skills without any demonstration and with only 10 minutes of real-
robot interaction. These results suggest that simulation can be a critical component in achieving physical intelligence, not only as a source of training data that must closely replicate the real world, but also as a roughly correct laboratory where a coding
agent develops the knowledge and procedures needed to act on the
robot. More details and
robot videos at https://robo-simex.github.io/
Read the full abstract ▾
All authors Jiaheng Hu, Roberto Martin-Martin, Peter Stone, Rocky Duan, Zhenyu Jiang, Guanya Shi
Distributed Computing
Vosti: Specifying, Implementing, and Verifying Deterministic LLM Inference
Jianxing Qin, Alexander Du, Danfeng Zhang et al. (+2 more)
LLM inference systems may vary batch composition,
prompt chunking, prefill/decode execution, and KV-cache reuse, eviction, or recomputation. These optimizations should not affect system outputs. Production systems, including vLLM's batch-invariant mode and SGLang's deterministic mode, target this goal but lack a formal system-level specification. We formalize deterministic LLM inference: under a fixed model and deployment configuration, requests with the same
prompt and initial sampler state produce bitwise-identical logits at corresponding output positions across executions. Our tests find that these production modes produce different logits under some execution variations. To address this limitation, we present Vosti, an inference engine designed and verified against this specification. Vosti chooses kernels independently of runtime engine state and ties cached KV values to their logical
token prefixes. Its proof decomposes at the engine/GPU kernel boundary: a Verus inductive proof establishes that scheduling and the paged, prefix-sharing KV-cache preserve output logits, while a Triton analyzer proves bitwise-equal selected kernel outputs across batches, query lengths, and paged KV-cache layouts. Vosti produces bitwise-identical logits across every tested execution variation and achieves performance comparable to vLLM's batch-invariant mode on decode-heavy workloads, while providing a stronger, formally verified determinism guarantee.
Read the full abstract ▾
All authors Jianxing Qin, Alexander Du, Danfeng Zhang, Matthew Lentz, Danyang Zhuo
Machine Learning
SCORE-LM: State-Space Radar Representations with Language Models for Fault Diagnosis
Mainak Mallick and Seung-Kyum Choi
Radar hardware faults threaten automated perception, motivating accurate, compact diagnosis and understandable maintenance guidance. We introduce SCORE-LM, which couples a small scatterer-conditioned operator-response encoder (SCORE) to an adapted local
language model. SCORE combines self-referenced complex trajectories, physical descriptors, and a selective state-space branch, with source-only self-supervision and directional fault inference. On eight capture-excluded Rad-R fault recordings, it achieves state-of-the-art performance within the evaluated nine-model comparison: 88.39% mean capture recall and 88.20% four-fault macro-F1 at ten frames. Its 39,520 radar inference coefficients are 119.7 times fewer than RadrNet-DS-CI's, while recall is 15.56 percentage points higher than this strongest competitor. In a separate low-label protocol, SCORE reaches 71.58% recall with one labeled source window per class. A nonlinear projector converts four frozen fault similarities into five soft
tokens, linking compact diagnosis to class-conditioned maintenance guidance. On 75 development questions covering 24 radar windows,
language adaptation raises correct-fault answers from 45 to 62 (60.0% to 82.7%) relative to removing the co-trained adapters, while retaining the same projector. SCORE-LM thus combines a compact radar specialist with a
language interface for communicating fault-specific inspection guidance.
Read the full abstract ▾
Computer Vision
Mitigating Object Hallucination in Large Vision-Language Models via False Discovery Controlled Visual Data Splitting
Chang Liu, Yu Tian and Rui Xie
Multiple object
hallucination, where large
vision-
language models (LVLMs) generate objects not supported by the visual input, is a persistent challenge caused by visual uncertainty during decoding. Existing methods reduce
hallucinations using contrastive signals, but they rely on heuristics and lack principled control of false positives at the image level. To address this, we propose False Discovery Rate-COntRol of
HALlucination (CORAL), a training-free framework that models visual uncertainty using an uncertainty-aware visual data splitting strategy and leverages mirror statistics to quantify visual contrast during decoding. By computing mirror statistics from paired, symmetrically perturbed visual inputs, CORAL estimates spurious object predictions and sets a data-driven threshold to control the expected fraction of false discoveries per image, suppressing
hallucinations while retaining high power for truly grounded objects. The framework is flexible, supports multiple LVLMs, and mitigates
hallucinations without retraining or supervision. Extensive experiments on multiple
benchmarks with several evaluation metrics demonstrate that CORAL consistently outperforms state-of-the-art methods, providing more reliable and robust
hallucination control. Code is available at: https://changliu1993-cl.github.io/CORAL/
Read the full abstract ▾
Computer Vision
Yun Dai, Jiarui Wen, Huiping Zhuang et al. (+2 more)
Diffusion Transformers (DiTs) have become a dominant architecture for
video generation, but their efficiency is limited by the quadratic complexity of full
attention. Sparse
attention reduces this cost by retrieving important blocks and computing
attention only within them, but inaccurate
retrieval can either degrade generation quality or yield unnecessary computation. We identify two
retrieval mismatches in methods that retrieve blocks using the averaged representations of query and key blocks: (i) query-side aggregation mismatch, where averaging queries before Softmax fails to preserve their individual
attention preferences, and (ii) key-side clustering metric mismatch, where standard Euclidean clustering in the original key space can group keys with dissimilar QK scores under the current query, so their average representation may not accurately represent how the current query scores individual keys. These mismatches can lead to inaccurate block
retrieval. To address these mismatches, we propose PARK, a training-free sparse
attention method for accurate block
retrieval. PARK retains every original query, independently normalizes its
attention over key blocks, and then averages these distributions within each query block. It also uses information from the current queries to transform keys before clustering, so that keys receiving similar QK scores are grouped together. A fused GPU kernel further reduces the overhead of block
retrieval. Experiments on HunyuanVideo and Wan demonstrate that PARK improves block
retrieval accuracy and preserves generation quality while accelerating inference, achieving the best quality-efficiency trade-off among the compared sparse
attention methods.
Read the full abstract ▾
All authors Yun Dai, Jiarui Wen, Huiping Zhuang, Cen Chen, Ziqian Zeng
Machine Learning
Scale-Split Neural Operator for Memory- and Data-Efficient 3D Turbulence Prediction
Shaoxiang Qin, Yucheng Zhao, Zongyi Li et al. (+2 more)
Neural surrogates have emerged as fast alternatives to the numerical simulation of three-dimensional
turbulence. However, training them at high resolution remains challenging, since the
memory of full-field models grows with the resolution. In addition, full-resolution training data are expensive to simulate and store, and therefore scarce. We introduce ScaleSplit-NO (Scale-Split Neural Operator), which exploits the scale structure of
turbulence with two neural operators: a Parent predicts the global coarse field at the next time step, and a Child predicts full-resolution local patches conditioned on this prediction. Neither model operates on the full-resolution field. The Child is pretrained alone and then attached to the Parent's coarse prediction through zero-initialized connections. On two complex high-resolution
turbulence benchmarks, ScaleSplit-NO surpasses all competing baselines in both prediction accuracy and data efficiency. On the higher-resolution
dataset JHTDB256 ($256^3$), its normalized mean squared error (NMSE) is 53% lower than that of the strongest baseline, and its training
memory is 79% lower than that of the most
memory-efficient baseline. We further demonstrate its effectiveness for urban wind prediction in a real district of Montreal on a $500\times150\times500$ grid, reducing one-step NMSE by 65.8% relative to the baseline. Moreover, swapping in a Parent trained on additional coarse fields improves prediction without retraining the Child, providing further accuracy gains at a small storage cost.
Read the full abstract ▾
All authors Shaoxiang Qin, Yucheng Zhao, Zongyi Li, Liangzhu Leon Wang, Xiongye Xiao
Computation and Language
Fairness Beyond a Single Run: Training-Seed Variability in Speech LLM Adaptation
Srishti Ginjala, Eric Fosler-Lussier and Srinivasan Parthasarathy
Demographic
fairness gaps in automatic
speech recognition are almost always reported from a single training run. We fine-tune the Q-former projector and LoRA adapters of a
speech LLM at five audio compression factors and six random seeds, holding the encoder, base decoder, data and decoding fixed, and evaluate every run on Common Voice and Fair-
Speech. At 460 h of clean LibriSpeech, the seed moves
fairness metrics more than compression does on most demographic axes. A balanced 3x3 decomposition attributes 85.3% of the variation in Fair-
Speech ethnicity normalized gap to the seed against 8.3% to compression (p = 0.009), though compression explains more on age and gender. Held-out LibriSpeech word error rate spreads by 0.04 points across those seeds while Common Voice spreads by 8.57, so these are not failed runs, and the effect survives controlling for accuracy and dropout. Scaling and diversifying the adaptation set to 960 h damps the effect but does not remove it. On Fair-
Speech ethnicity, two single-run systems must differ by more than 0.30 in normalized gap to exceed seed variability.
Read the full abstract ▾
Artificial Intelligence
RealWorldShop: Benchmarking and Improving Conversational Shopping Agents in Real-World E-commerce
Xinwei Yang, Kelong Mao, Yudong Guo et al. (+4 more)
Large
language models are reshaping ecommerce from static recommenders into interactive shopping assistants, yet real-world shopping requires session-level decision support: users reveal and revise constraints, coordinate multiple goals, and expect product-grounded recommendations over a full conversation. Existing
benchmarks are mostly outcome-oriented or execution-oriented, leaving this evolving decision process under-evaluated. We introduce REALWORLDSHOP, a
benchmark built on 3.28M grounded products, structured shopping episodes, a profile-grounded and actioncontrolled user simulator, and role-play evaluation. Our analysis shows that current systems produce locally plausible responses but struggle with state tracking, constraint updating, and grounded convergence, especially under ambiguous intent, bundle, and multi-intent scenarios. We further propose REALSHOP_
AGENT, an executable session-control framework with explicit state management, shopping-flow control, catalog-grounded
retrieval, and runtime guards. Experiments show that REALSHOP_
AGENT consistently outperforms strong baselines on REALWORLDSHOP.
Read the full abstract ▾
All authors Xinwei Yang, Kelong Mao, Yudong Guo, Sulong Xu, Simiu Gu, Chen Huang, Wenqiang Lei
Computation and Language
Making LLMs Say What They Think: Measuring and Improving CoT-Interpretability Alignment
Yihuai Hong, Shauli Ravfogel, Chen Zhao et al. (+1 more)
Chain-of-thought (CoT) traces often serve as a proxy for how Large
Language Models (LLMs) arrive at their answers. However, growing evidence shows that models' CoT often fails to reflect their internal computations and can be changed without affecting their final answers. In this work, we measure and improve the
alignment between the
reasoning described in an LLM's CoT and what it computes internally. We propose CoT-Interpretability
Alignment (CIA), a metric that measures the agreement between a model's CoT traces and its internal
reasoning strategies as detected by interpretability tools. We evaluate CIA on three tasks (two-hop question answering, hint intervention, and integer multiplication) across three LLMs, finding that LLMs exhibit limited
alignment across all tasks (44.8-75.9%). We then experiment with improving CIA via post-training, setting both the task accuracy and parametric faithfulness signals as a reward. Experiments show that we can substantially improve CoT parametric faithfulness while maintaining or improving the task accuracy. We provide rich analysis, such as their generalization patterns. Our work provides both a framework for
auditing CoT parametric faithfulness and a pathway toward making models' explicit
reasoning more trustworthy. Code and data are available at https://github.com/yihuaihong/CIA-minimal-repro.
Read the full abstract ▾
All authors Yihuai Hong, Shauli Ravfogel, Chen Zhao, Eunsol Choi
Cryptography and Security
Refusals That Bend: Measuring and Predicting Task Malleability in Embodied VLM Planners
Leo Y. Lin, Mikhail Kuznetsov, Muslum Ozgur Ozmen et al. (+1 more)
Embodied vision-
language models (VLMs) are increasingly deployed as high-level planners for
robots because they generalize across diverse environments. However, this requires their safety
alignment to also hold in unseen environments. Existing red-teaming assumes an adversary who optimizes the
prompt, the pixels, or text in the environment, and existing
benchmarks ask whether a planner recognizes or mitigates a hazard in a fixed scene. Neither asks whether a refusal the planner has already given survives an ordinary change to the environment. We ask that question by placing a single everyday object into the environment, with no pixel, gradient, or
prompt under
adversarial control. On $846$ tasks that a constitution-guarded planner initially refuses, we find $20.2\%$ of tasks can be flipped to compliance by one or more objects, and the number of objects differs from one task to another. In addition, the object need not be chosen for the task, i.e., items drawn from a fixed list, with no knowledge of the environment or the instruction, bypass safety about as often as items proposed for the specific task. We qualitatively contrast the tasks bypassed most and least often and find that the distinction lies in how conspicuous the hazard is in the instruction and environment. Susceptibility to safety bypass is therefore a property of the task, which we call its malleability, and we show that it can be predicted before the target is ever queried. A composite of signals read from a small open-source VLM identifies malleable tasks $2.4\times$ as often as picking at random. Everyday objects, whether placed by an adversary or introduced by ordinary rearrangement of the environment, are thus sufficient to overturn a refusal. Because susceptibility is determined by how a task is specified, we recommend assessing malleability per task prior to deployment.
Read the full abstract ▾
All authors Leo Y. Lin, Mikhail Kuznetsov, Muslum Ozgur Ozmen, Z. Berkay Celik
Human-Computer Interaction
Insights on Student Learning from Live Classroom Polls: More Than Right or Wrong
Rohit Sharma, Pavani Ayinampudi, Aditya B.M.V. et al. (+5 more)
A live classroom poll is usually read through its answer key, as a count of participation and correctness. Yet the same record holds more before any key is fixed, namely who answers early and how the class divides across the options. Whether these key-free signals are stable, and what they reveal, has not been examined at the scale of a real course. We study them in a live deployment of 340,668 submissions across 603 polls in 47 sessions from roughly 2,400 learners, reading each poll for a learner's response order and for how much the class divides. A single further step brings in the author-designated key, on items with a settled answer, to interpret the divided ones. A learner's response order is a stable individual signature. It reproduces at a corrected split-half reliability of 0.92, holds at 0.87 within single sessions, and is unrelated to whether the learner is correct. Read without a key, about one poll in five does not reach a clear majority, and division marks the questions that split the class. On the settled-answer items the majority is itself wrong on 49 of 358. The key-free measure flags most of these, though division alone does not separate a collectively wrong answer from a merely hard one. The two readings are largely independent, meeting at one point where early responders anticipate the eventual majority only on questions the class has nearly settled. Together they map the learners and questions of a session without a key.
Read the full abstract ▾
All authors Rohit Sharma, Pavani Ayinampudi, Aditya B.M.V., Jinal Gupta, Prakash Hegade, Sakshi Sharma, Meenakshi V, SRS Iyengar
Computer Vision
Beyond Spatial-Domain Supervision: A Relation Constrained Space for Multi-Modal Image Fusion
Zeyu Wang, Mingyu Ge, Haiyu Song et al. (+1 more)
Multi-modal image fusion (MMIF) aims to form a single image by integrating shared information, preserving complementary cues, and coordinating cross-modal conflicts across modalities. However, due to the absence of ground-truth fused images, existing MMIF supervision commonly uses spatial-domain sources or gradient variants as surrogate ground truth, making the supervision mechanism inherently misaligned with the goal of MMIF and causing pixel-level compromise or modality bias. To address this, we propose a relation-constrained supervision paradigm that moves fusion supervision from the spatial domain to a learned relation space. Rather than relying solely on direct source approximation, we further leverage frozen pretrained representation models as information providers and design a learnable feature adapter to align heterogeneous DINO and CLIP features into a unified supervision space. The adapter infers three relation parameters, namely sharedness, dominance, and coordination radius, which define three losses corresponding to the MMIF's goal. To make this space reliable, we devise a self-supervised contrastive ranking objective tailored to the adapter and couple it with the fusion network through alternating optimization. Extensive experiments show that the proposed supervision space yields significant gains regardless of which mainstream backbone the fusion network adopts, offering a supervision paradigm better aligned with the goal of MMIF. Code: github.com/GMY628/RCS-Fusion.
Read the full abstract ▾
All authors Zeyu Wang, Mingyu Ge, Haiyu Song, Haoran Duan
Human-Computer Interaction
What Students Actually Ask: Demand Structure and Automation Potential in a Hybrid Support System
Jinal Gupta, Pavani Ayinampudi, Aditya B.M.V. et al. (+5 more)
Large online programmes receive heavy volumes of queries during onboarding, at a scale that grows faster than the number of staff available to answer them. This paper reports a study of an AI integrated query resolution platform that spreads incoming queries across four routes: an AI based assistant, answers from fellow participants, a curated corpus of frequently asked questions, and escalation to administrators. Over nine weeks, from 2 May to 6 July 2026, the system handled 4,093 queries raised by 1,434 participants. Nearly every query reached a recorded resolution, and one query in five closed within an hour. Reuse of 114 corpus entries absorbed 21.3% of the volume, participants resolved a further 12.2% on their own, and 132 participants answered questions for one another at a median of 9 to 15 minutes, showing that peer answering, where it occurred, was fast and broadly shared across the cohort. Classifying the query text shows that demand was narrow rather than varied. A single process step, the submission of a certificate and the offer letter that follows it, accounts for 56.8% of corpus mediated resolutions, and at least 20.4% of queries concern the progress of a pending submission rather than a request for information, a class the assistant served only 1.3% of the time, since a stored answer cannot report an individual's current status. Only 8.0% of the queries handled by a person duplicated content already in the corpus, which indicates that the knowledge base was already well used. Most direct administrative closures occur in synchronous bulk events, a pattern that shapes how the records should be read.
Read the full abstract ▾
All authors Jinal Gupta, Pavani Ayinampudi, Aditya B.M.V., Prakash Hegade, Rohit Sharma, Sakshi Sharma, Meenakshi V, S.R.S. Iyengar
Robotics
Video2SwimFish: An Automated Pipeline for Reconstructing Controllable Fish Models and Biological Locomotion from Real Fish Videos
Hangong Chen, Linfeng Cheng, Tahsin Zaman Jilan et al. (+4 more)
We present
Video2SwimFish, an automated pipeline and
benchmark for building controllable fish assets from real-fish
videos for
underwater embodied AI. Given synchronized multi-view
videos of an individual fish, the pipeline reconstructs a metrically scaled deformable mesh from a VLM-selected canonical frame, generates internal articulation adapted to that individual's morphology through a VLM actor-critic loop, and extracts a Biological Locomotion
Manifold (BLM) from the fish's observed midline
curvature. The BLM provides a low-dimensional action space bounded by real-fish motion, enabling an individual swimming policy to be learned for each reconstructed fish. We release two paired
datasets: synchronized top- and front-view recordings of 120 individual fish across 6 species, and the controllable assets and individual swimming policies derived from them. Because every asset is tied to the animal it came from, the
dataset supports a
benchmark that evaluates locomotion learning not only on task success but on fidelity to that individual in trajectory shape, body
curvature, and tail-beat frequency, across trajectory following, reward-free swimming behavior transfer from
video, and a downstream case study in which a simulated BlueROV
underwater robot captures one of the assets. We find that task success and locomotion fidelity do not necessarily improve together: the method achieving the highest task completion is not the method achieving the highest locomotion fidelity, and we identify faithful reproduction of individual animal locomotion as an open challenge for the community. Project website: https://hangongchen.github.io/
video2swimfish-web/.
Read the full abstract ▾
All authors Hangong Chen, Linfeng Cheng, Tahsin Zaman Jilan, Ian Fuller, Lee Caesar, Jiaye Wu, Yantian Zha
Numerical Analysis
A positive-conservative Fourier optimization method for ground states of Bose–Einstein condensates with higher-order interactions
Bo Lin and Xinran Ruan
We develop a positive-conservative
Fourier optimization (PCFO) method for computing ground states of Bose–Einstein condensates with higher-order interactions. The ground-state problem admits a convex density formulation, but the singular behavior near zero density poses difficulties for high-accuracy computation. We introduce convex regularizations of the density energy and establish their $\Gamma$-convergence on the computational domain as the regularization parameters vanish. The regularized problem is discretized by a
Fourier pseudospectral method with mass conservation and nodal positivity. For the resulting constrained optimization problem, we use an accelerated first-order method for the main energy reduction, followed by a Newton refinement to reduce the remaining optimality residual. Numerical experiments show spectral-type convergence for regularized problems, quantify the effects of the regularization parameters on accuracy and spatial resolution, and demonstrate the effectiveness of the two-stage optimization method.
Read the full abstract ▾
Artificial Intelligence
When Order Matters: First-Speaker Bias and Mitigation through Personality in Sequential Multi-Agent Debate
Duofeng Xu, Bryan Hooi and Dandan Qiao
Multi-
agent debate (MAD) is often used to improve large
language model (LLM)
reasoning, but sequential debate is rarely a neutral aggregator of
agents' opinions. We show that sequential MAD suffers from a pronounced first-speaker bias:
agents disproportionately shape the final answer when they speak first. As a result, placing a stronger model after weaker ones can substantially offset its
reasoning advantage. We then focus on the disadvantaged strong-
agent-last setting and ask whether personality prompting can mitigate this imbalance. Drawing on the Big Five model, we study agreeableness and extraversion as behavioral interventions applied to either the strong or weak side. We find that their effects are trait-specific. Influence consistently shifts in the direction of lower agreeableness, and assigning low agreeableness to the stronger
agent helps restore its lost influence and improves final accuracy. Extraversion, by contrast, produces less systematic changes in influence and accuracy, with its clearest effect appearing in
agents' verbosity. These findings show that effective MAD design depends not only on model capability, but also on how speaking order and induced interaction behavior shape the debate process.
Read the full abstract ▾
Artificial Intelligence
Alleviating Hallucination in Reasoning Tasks with Training-Free Uncertainty-Guided Steering
Litian Liu, Qiqi Hou, Yubing Jian et al. (+5 more)
Recent work on
hallucination detection in large
language models has shown that, for a fixed pre-trained model and
reasoning task, it is possible to estimate the model's confidence in the correctness of its outputs. Such uncertainty estimates have primarily been used to improve truthfulness by detecting or filtering confabulations. In this work, we ask whether these signals can instead be used more proactively to directly improve the accuracy of model-generated answers. We propose USteer, a simple, training-free steering mechanism that adjusts a model's layer-wise activations during inference using the gradient of a confidence measure with respect to the activations. This procedure nudges generation toward outputs with lower uncertainty at inference time, without modifying model parameters or requiring additional supervision. We show that this approach consistently reduces
hallucination across a range of tasks, demonstrating that confidence signals can be leveraged not only for detection, but also for effective inference-time control of model behavior.
Read the full abstract ▾
All authors Litian Liu, Qiqi Hou, Yubing Jian, Reza Pourreza, Mohammad Ghavamzadeh, Roland Memisevic, Yao Qin, Hong Cai
Multiagent Systems
Fast and Scalable Multi-Agent Distribution Matching via Partitioned Optimal Transport
Kooktae Lee and Ruchika Singh
This paper presents a scalable optimal-transport-based framework for terminal distribution matching in multi-
agent systems. While optimal transport provides a natural way to measure distributional mismatch and assign
agents to a desired spatial distribution, global discrete transport can become computationally expensive for large-scale systems. We address this bottleneck by partitioning
agents and target samples into spatially corresponding blocks and solving smaller local transport problems. Under a mass-balance condition, the resulting restricted coupling remains feasible for the global problem and provides an upper bound on the Wasserstein cost. The local assignments generate target locations for finite-horizon
agent control, applicable to both linear and nonlinear dynamics. By alternating local assignment and control, we establish a cycle-to-cycle descent guarantee for the resulting transport surrogate. The proposed framework therefore enables scalable terminal distribution matching while retaining a rigorous connection to the Wasserstein objective. The technical soundness of the proposed results is validated through simulations.
Read the full abstract ▾
Human-Computer Interaction
An Audit of Measurement Quality and Answer Bias in a Large Classroom-Poll Corpus
Rohit Sharma, Pavani Ayinampudi, Aditya B.M.V. et al. (+5 more)
Real-time classroom polls are widely used and increasingly generated with automated assistance, yet the questions themselves are rarely evaluated as measurements. We audit a large corpus of authentic classroom polls, 604 items across 47 sessions answered 340,668 times by 2,807 learners, as a measurement instrument. For the 539 items whose correct answer could be established and verified from the lecture transcript, we place every item and every student on a common scale using item response theory and analyse the answer structure of the True/False items. Two findings emerge. First, the polls form a coherent but easy scale of moderate precision (marginal reliability about 0.60), on which roughly a quarter of items barely separate stronger from weaker students. Second, students show a robust tendency to answer True, present at the individual level (77% of students lean True), which meets a milder tendency for items to be keyed False; as a result answer direction predicts difficulty, False-keyed items being about thirteen points harder, and the effect survives controls for item content and for selective answering. Both findings rest on signals a polling system already records, so the same checks can be run as items are generated, before they reach students.
Read the full abstract ▾
All authors Rohit Sharma, Pavani Ayinampudi, Aditya B.M.V., Jinal Gupta, Prakash Hegade, Sakshi Sharma, Meenakshi V, SRS Iyengar
Artificial Intelligence
Targeted Retrieval, Compact Representations: How CoT Reasoning Improves Long-Context Counting
Liang Twist Shan, Tianyu Hu, Hao Yan et al. (+1 more)
Large
language models (LLMs) have been rapidly improving in long-context tasks, powered by Chain-of-Thought (CoT)
reasoning. However, the internal mechanisms underlying this improvement remain unclear. We investigate these mechanisms through a needle-in-a-haystack (NIAH) counting task, where an LLM is asked to count the number of records dispersed in a long text. Across twelve model comparison groups, Thinking (or
reasoning) improves counting accuracy over Non-thinking, with pronounced gains at larger counts. This motivates our mechanistic analysis, which identifies two contrasting mechanisms: (i) broad
retrieval, where Non-thinking models broadly attend to multiple needles; (ii) targeted
retrieval, where Thinking models use enumeration in CoT traces to successively retrieve needles. Targeted
retrieval concentrates
attention on individual needles and is accompanied by more compact internal representations. Moreover, causal intervention analysis suggests that Thinking models use the CoT trace to maintain and update an internal counter as needles are successively retrieved, even without explicit numbering. In small controlled experiments, both
retrieval mechanisms and counter states emerge under standard autoregressive training. Together, our results connect long-context
retrieval with representation geometry of counting, supporting a state-tracking account of CoT
reasoning.
Read the full abstract ▾
All authors Liang Twist Shan, Tianyu Hu, Hao Yan, Yiqiao Zhong