Ziyi Luo, Zhe Sun, Yehao Lu, Lei Zhou, Lisheng Wu, Xuewei Li · Sep 29, 2026 · Citations: 0
Automatic MetricsGeneral
Average performance on routine driving benchmarks does not establish planner reliability under rare, safety-critical hazards.
We proposed ExceptionDrive, a counterfactual planning benchmark that uses VLM-assisted screening, localized multi-view editing, and quality auditing to insert hazards into real nuScenes scenes while preserving their context.
On the LACUNA localization-precision benchmark, our mean terminal utility is higher in all six negative preference optimization (NPO) and SimNPO comparisons: NPO margins range from +0.431 to +0.848, and SimNPO margins range from +0.503 to…
Relative to Static-IV, the primary four-field GradDiff evaluation has six wins, six ties, and no losses, with mean and median paired gains of +0.165 and +0.0025.
Haotian Deng, Wenbin Xing, Gang Xu, Tao He, Jinkai Zheng, Chun Li · Sep 29, 2026 · Citations: 0
Pairwise PreferenceAutomatic MetricsGeneral
Existing safety methods often require large preference datasets, costly multi-rollout training, or additional safeguards at inference time.
With only 1,447 safety-specific examples - 95% fewer than standard preference datasets - OPSD reduces training time by 5x relative to multi-rollout GRPO-style training and average inference length by 7%.
Serafima Lebedeva, Sumantrak Mukherjee, Ali Arshad Sadal, Ilias Ekşi, Rahul Sharma, Julia Mueller · Sep 29, 2026 · Citations: 0
Pairwise PreferenceSimulation EnvGeneral
This creates two challenges: learning user preferences quickly from limited feedback and sustaining useful recommendations when each user has a finite catalog that can become repetitive or depleted over time.
Our evaluations show that cross-cohort transfer improves early recommendation quality and user-level regret, while inventory-aware slate construction helps prevent premature exhaustion of preferred items.
This allows the model to learn what is most important to maintain in context, and naturally extends to multi-agent systems where multiple agent contexts coexist as files.
Building CLMs zero-shot with existing models outperforms SOTA context management strategies across a variety of tasks: 11.4% higher accuracy with 21.5% fewer FLOPs on BrowseComp-Plus, 5% higher scores with 59% fewer FLOPs on 12-hour…
Extensive experiments across OCRBench v2.1, CC-OCR, in-house KIE Benchmark, OmniDocBench v1.6 and MDPBench demonstrate that PolyOCR achieves state-of-the-art or highly competitive performance.
T. Duy Nguyen-Hien, Yee Whye Teh, Wee Sun Lee, Tan Zhi-Xuan · Sep 29, 2026 · Citations: 0
Pairwise PreferenceGeneral
Users of language-based assistive agents often make ambiguous requests.
In two assistive tasks --- ambiguous question answering (CondAmbigQA) and preference-aligned household task planning (ADAPT) --- we show that REVOIR achieves greater success with fewer questions than approaches based on prompting,…
Unlike APO methods that judge candidates with an LLM evaluator, our loop scores each candidate by its downstream retrieval performance and runs a small tournament between the current champion prompt and optimizer-proposed drafts,…
On five BEIR benchmarks (NQ, SciFact, FiQA, Touche-2020, DBPedia), MERGE improves BM25 nDCG@10 over the original queries by +2.1 to +14.9 points and matches or outperforms strong LLM-based query-expansion baselines despite using only…
Long-horizon information-seeking agents often accumulate noisy or misleading context, causing early mistakes to persist and making recovery increasingly difficult.
We introduce an autonomous search harness in which the agent manages its own search process through three states: Rubric, Answer, and Verify.
We propose LatCom, a cross-agent latent compression framework for efficient multi-agent latent collaboration.
Experiments on multiple benchmarks with Qwen3-4B show that LatCom achieves an average 2.46x inference speed-up over LatentMAS and reduces output token usage by 70.3% while maintaining comparable average accuracy.
In this paper, we propose a novel three-stage Select-Reason-Judge (SRJudge) framework, which empowers LLMs with selective reasoning capability for fine-grained knowledge concept tagging.
It further integrates an improved reinforcement learning strategy with a dynamic task-specific reward function and a pruning mechanism to better align with human reasoning preferences.
The controller freezes its decision and cost ledger before joining the clean oracle; a dependence-shift alarm disables channel preference and falls back to exact-stop.
End-to-end full-duplex speech models have brought open-source machine conversation closer to human-like interaction, yet existing systems remain limited in two intertwined dimensions: long-context robustness and multi-party interaction.
Second, we introduce https://huggingface.co/datasets/MultiTalk/MultiTalkBench{MultiTalkBench}, built from real human recordings, for evaluating long-form, multi-party, bilingual full-duplex dialogue.
Some dynamic rubric methods alternate updates between the caption policy and rubric generator while keeping the judge fixed, but staged optimization may still leave rubric construction and judging out of step with policy optimization.
Across five captioning benchmarks, MoCo Rubric achieves an average pairwise win rate of 72.83\%, the best mean rank in blind ranking, and the highest average score in caption-based question answering.
Personal agents bring this question into everyday settings, where models are expected to serve individual users and continually adapt to their preferences.
Empirically, we introduce a preference-oriented benchmark and systematically characterize the capabilities and limitations of personal agents associated with these three dimensions.
Sen Zhao, Ruiqi Kong, Zuyu Zhang, Lifeng Shen, Xinyu He, Ding Zou · Sep 26, 2026 · Citations: 0
Automatic MetricsGeneral
Multi-agent workflows require task-relevant information to be shared across agents, while irrelevant, stale, unverified, or incompatible information must remain isolated.
We introduce CoMemBench, an execution-grounded benchmark for collaborative memory sharing and isolation across multi-agent workflow topologies.
Recursive self-improvement is becoming essential for autonomous AI agents, whose progress depends on discovering high-value solutions across complex domains.
A lightweight orchestration layer makes exploration explicit and programmable while leaving the underlying base agent unchanged.
