The paper introduces LibStructBench, a document‑grounded benchmark designed to assess whether AI agents can accurately reconstruct sequencing‑library structures and molecular workflows from protocols. Experiments reveal that current state‑of‑the‑art agents often generate roughly correct reconstructions but struggle with exact molecular‑state representation and workflow connectivity. The study also finds that changing the underlying language model yields larger performance differences than changing agent harnesses, and that additional knowledge only modestly improves oligo‑family recovery without affecting overall workflow accuracy.
By Chi-Lam Poon, Michael Goneos, Zezheng Song, Xi Chen
The paper presents a closed‑form model that compares the coordination tax of multi‑agent LLM pipelines—comprising decomposition overhead and per‑handoff protocol cost—to the heterogeneity dividend gained by routing work to cheaper models. Using the Open Floor Protocol (OFP) on 135 HumanEval and ClassEval tasks, the authors find that the protocol adds only 7.2 ms per handoff, yet a planner‑coder‑reviewer pipeline is 3.22× more expensive and slower than a single strong model call, and does not solve more tasks. The study concludes that the benefit lies in selecting the appropriate model per task rather than in pipeline decomposition, as an oracle choice between two single calls solves 94.3 % of runs at half the cost of the strong call.
By Andreas Zettl, Diego Gosmar
The paper investigates positional bias in large language model (LLM)–based recommender systems that use retrieval‑augmented generation (RAG). It introduces two statistical detection methods—POB‑ICM and POB‑UNBP—to test whether the recommended item depends on the order of candidates in the prompt. Experiments on a product‑recommendation testbed with 40,000 instances show that all evaluated state‑of‑the‑art models exhibit significant positional bias, quantified by new metrics such as permutation sensitivity and strong positional bias, and that this bias is structural and model‑dependent rather than an artifact of decoding or candidate‑set size.
By Carlos Pel\'aez-Gonz\'alez, Francisco Herrera, Andr\'es Herrera-Poyatos
The paper introduces Proof‑Carrying Messages (PCM), a protocol that attaches verifiable evidence—Merkle log records and signed attestations—to each claim exchanged between AI agents. PCM enables receivers to evaluate claims offline, categorizing them as proven, unproven, or failed, and ensures that claims based solely on an agent’s own report can never be proven. In benchmark tests on 926 real messages, PCM accepted only 1.4% of false claims, whereas other protocols accepted many more.
By Bravish Ghosh (Cygnux Labs)
The paper introduces Learning to Bound (L2B), a data‑driven framework that trains a neural predictor to produce admissible, monotonic upper bounds for Most Probable Explanation (MPE) inference in probabilistic graphical models. L2B achieves admissibility through statistical calibration and ensures progressive tightening via architectural constraints that enforce monotonicity. Experiments on challenging benchmarks show that L2B delivers tighter, monotonic bounds at orders‑of‑magnitude lower cost than classical methods and improves branch‑and‑bound performance on high‑treewidth instances.
By Akshay Vyas, Hailiang Dong, Shivvrat Arya, Brij Malhotra, Tahrima Rahman, Vibhav Gogate, Nicholas Ruozzi
The paper introduces ExLabel, an LLM‑based, self‑evolving assistant that assesses task clarity and test coverage in SWE‑bench‑style software engineering datasets. ExLabel builds an experience bank from cross‑instance knowledge, enabling context‑aware judgments that outperform static rubrics. A new benchmark, SWE‑TQ, spanning eight programming languages, demonstrates ExLabel’s superior F1 scores and its ability to filter data that improves a coding model’s success rate on SWE‑bench Verified.
By Zhonghao Jiang, Xuyang Zheng, Yanfu Yan, Zhongxin Liu, Lingfeng Bao, Shuiguang Deng
The paper introduces Task-Graph Contextual Integrity (TG‑CI), a formalism that extends traditional Contextual Integrity to multi‑agent systems by modeling workflows as Multi‑Agent Task Graphs (MATGs). TG‑CI evaluates privacy not just at individual disclosures but across the entire downstream propagation, offering two enforcement policies—Full‑Scope and Auto‑Region—that balance privacy protection with task performance. A new benchmark, MAPLE‑Bench, demonstrates that TG‑CI can cut unauthorized disclosures by up to 94% while preserving over 94% of task accuracy compared to existing local‑history CI methods.
By Hengyu Wu, Blaise Delattre, Yang Cao
arXiv:2610. 39676v1 Announce Type: cross Abstract: Event cameras offer microsecond temporal resolution, high dynamic range and no motion blur, but referring video object segmentation (RVOS) lacks event-stream datasets and models.
By Sergei Fedorchenko, Max Zehnder, Nikola Zubi\'c, Rong Zou, Davide Scaramuzza
The paper evaluates how large language models, when integrated into a harness that provides information retrieval and software execution, perform as agents in acoustic engineering tasks. In one experiment, the agent accurately detects and locates faults in vehicle interior audio recordings, showing that a weak model limits performance regardless of harness design. In a second case study, the agent reliably designs metamaterial unit cells to meet specified insertion loss targets across a frequency range, illustrating the agent’s practical utility in acoustic design.
By Arne Hildenbrand, Philipp Diel, Zhuo Cao, Hanno Scharr, Stefan Pischinger, Steffen Marburg
The paper introduces SIFR, a frequency-aware dual‑stream framework for domain‑generalized semantic segmentation. SIFR performs spectral‑guided invariant fusion of frozen CLIP and DINOv2 features to suppress domain‑sensitive responses while preserving semantic and structural cues, and adds a spectral decoder refinement module that normalizes low‑frequency statistics and adaptively selects frequencies to reduce residual domain bias. Experiments on synthetic‑to‑real, real‑to‑real, and adverse‑weather benchmarks show consistent improvements over prior state‑of‑the‑art methods.
By Yansong Bai, Yunzhi Zhuge, Xinzhuo Yu, Lu Zhang, Jin Zhan, Pingping Zhang, Huchuan Lu
The paper introduces LoCo, a model-agnostic framework that localizes where and when specific concepts appear in visual autoregressive models across Layer, Scale, and Position dimensions. It also presents LoCoBench, a dataset covering ten categories, and demonstrates the method on models such as Infinity, HunyuanImage-3.0, and InfinityStar. Using these localizations, the authors apply concept erasure, model personalization, and adversarial concept injection, achieving state‑of‑the‑art performance with reduced computational cost.
By Nanxiang Jiang, Yang Liu, Yuanhao Wang, Min Xu
The paper introduces Branch-aware Graph Planning (BGP), a method for offline goal-conditioned reinforcement learning that tackles the problem of path mirages in stochastic, long-horizon environments. BGP models stochastic branching by treating high-variability points as graph nodes and learns goal-conditioned edges with segmented learning to avoid unreliable branches and reduce unnecessary stochasticity. Experiments on diverse OGBench tasks demonstrate that BGP substantially outperforms existing state-of-the-art offline goal-conditioned RL approaches.
By Junhyeon Bae, Jaebak Hwang, Gwanwoo Choi, Minung Kim, Seungyul Han
The paper introduces a unified Hessian-guided framework for post‑training quantization of large language models. By deriving a Hessian‑based approximation of output changes, it jointly optimizes quantization ranges, linear transformations, and weight rounding under a single output‑sensitivity objective. The method remains practical for modern models and consistently outperforms existing transformation and rounding techniques in both channel‑wise and block‑wise settings.
By Mads Toftrup, Marco Federici, Boris van Breugel, Paul Whatmough, Markus Nagel
The paper introduces IMPACT, a physics‑informed multimodal deep‑learning framework that forecasts photovoltaic power 10–110 minutes ahead with probabilistic outputs. It uses a four‑branch architecture to encode plant‑level temporal history, inverter‑level spatial cloud dynamics, known‑future solar geometry, and current meteorological data, fusing them via gated cross‑attention. The model outperforms 14 baselines, achieving a forecast skill score of 0.2950 and maintaining strong performance under overcast and variable sky conditions.
By Anushka Katiyar, Cong Feng, Jing Wang, Tristan Liu, Simon Julien, Zachary Jacobs
BURGER is a benchmark for evaluating vision‑language‑action models on robotic drawing tasks. It provides a scalable, prompt‑driven pipeline that generates continuous drawing trajectories in both simulation and the real world, with 28,200 simulated demonstrations across 282 tasks. The benchmark enables fine‑grained analysis of language grounding, spatial composition, and action generation, revealing distinct failure patterns among five representative VLA models and demonstrating that some models can generalize to unseen compositions and scale with task diversity.
By Xiangyu Xu, Bangshuai Peng, Ziyi Liu, Mengjie Zhao, Chenhao Lu, Zhengyang Fan, Huazhe Xu, Kaizhe Hu
PRISM is a framework for proactive agents that decides when to intervene by using a cost‑sensitive decision gate and a dual‑process reasoning architecture. The agent only engages a resource‑intensive slow reasoning mode near the decision boundary, guided by a calibrated probability of user acceptance and asymmetric costs for missed help versus false alarms. Training employs gate‑aligned distillation, allowing the response policy to be decoupled from the gate for tunable, auditable control, and on the ProactiveBench benchmark PRISM reduces false alarms by 22.78% while improving F1 by 20.14% over strong baselines.
By Yuxuan Fu, Xiaoyu Tan, Teqi Hao, Chen Zhan, Xihe Qiu
MuQ-Eval is an open‑source per‑sample quality metric for AI‑generated music that uses frozen MuQ‑310M features and a lightweight prediction head trained on the MusicEval dataset. The simplest model—attention‑pooled frozen features plus a two‑layer MLP—achieves high correlation with human judgments (system‑level SRCC = 0.957, utterance‑level SRCC = 0.838). The metric is real‑time on a single consumer GPU, outperforms existing open per‑sample metrics, and supports personalized evaluators via LoRA adaptation on as few as 150 clips.
By Di Zhu, Zihan Chen, Zixuan Li
Hedge-Bench is a benchmark comprising 92 real-world environments derived from hedge fund analysts' reasoning traces, designed to test AI agents’ ability to forecast future events from incomplete information. Agents are evaluated on joint success—correct forecasts with over 60% research coverage—across 11 frontier models, achieving rates between 2.54% and 31.52%. Training Muse‑Glimmer with GRPO on 165 environments improves coverage on held‑out tasks and demonstrates generalization to out‑of‑distribution evaluations.
By Mahesh Ramesh, Eric Cho, Alice Lu, Shawn Huang, Baiqing Lyu
RePolicy is a safeguard that learns to both select the appropriate safety policy from a library and assess agent behavior against it. It processes an agent trajectory and a set of candidate policies, chooses the best-matching policy, applies its requirements, and predicts safety. The authors created PolicyTraj-20K, a dataset of over 20,000 annotated trajectories, and trained RePolicy with supervised and reinforcement learning, achieving top performance on six benchmarks and 19 baselines.
By Houcheng Jiang, Boxuan Zhang, Qiyong Zhong, Junfeng Fang, Xiang Wang, Xiangnan He
SemantiScene-Edit is a video editing system that uses a multimodal encoder and a pretrained video diffusion transformer to follow natural‑language instructions while preserving the source video’s appearance and temporal structure. The encoder processes source frames and the instruction together, sending semantic queries to the generator via a trainable connector and cross‑attention. A dual‑path visual route concatenates source‑video latents with noisy target latents to provide dense spatial and temporal information, and the model is trained with flow matching and a three‑stage curriculum on over 1.2 million instruction‑conditioned records, with a 100‑case benchmark for evaluation.
By Xinyu Chen, Song Wu, Dingbao Shao, Jiahang Li, Jiang Lin, Yuhang Liu, Duo Li, Jiaxin Hu, Shengrong Gu, Liang Li, Qian Wang, Junlan Feng, Ying Tai, Jian Yang, Zili Yi