The effectiveness of a self-improving system relies heavily on the quality of its verifier, which judges the outcome of each improvement round. However, the verifiers of most tasks are written by hand, which is expensive and does not scale to tasks that come without one. We propose Veritas, a harness that generates ver...
Dao-An Zhang, Qin-Bo Bai, De-Hai Min et al.· 0 citations
VerifierBench, a benchmark of 222 tasks from six agent benchmarks spanning tool-call, state-based, and code verification, with 4,993 hidden test cases that include alternative valid solutions and attacks is introduced; Verifier4Verifier (V4V), a six-check audit, screens its source verifiers and test suites.
Jia-Ning Yin, Zi-Yi Chen, Hao Wang et al.· 1 citation
This work proposes RLiD (Reinforcement Learning in the Decision plane), a framework that directly optimizes the probabilities of executable action options and avoids invalid action outputs and improves RL rollout efficiency by eliminating autoregressive token generation and bypassing action parsing.
Zi-Yi Chen, Yan Zhang, Jian-Hui Wei et al.· 0 citations
Thought templates, skill libraries, agent memories and caches turn many reasoning steps into one. We call this reasoning contraction. A contracted memory saves reasoning steps and errors until its facts change, then quietly answers wrong. Most systems trust memories by counting successes and failures. We prove that suc...
Jiang-Hong Zhou, Hao-Yan Rong, Chen-Ying Lin et al.· 0 citations
Self-Posed Reasoning Distillation (SPRD) adds tasks generated by a frozen teacher, filters them for answer consistency, and trains on label distributions obtained from repeated teacher reasoning, achieving a 182-fold speedup in latency over reasoning rollout on Gemma-4-31B.
Zi-Yi Chen, Yan Zhang, Jian-Hui Wei et al.· 1 citation
Words are increasingly becoming an interface for action: large language models (LLMs) can now invoke tools, operate devices, and make decisions that affect the physical world. However, current benchmarks largely evaluate LLMs through static text responses, offline tool calls, or simulated tasks, leaving open the questi...
Xin-Nan Dai, Kai-Qi Yang, Ya-Xing Li et al.· 0 citations
When training large language models with reinforcement learning, terminal rewards provide little guidance about which steps matter. Common methods for assigning step credit overlook that work built on uncorrected mistakes is wasted while independent work remains valid. With only a final success/failure reward, every st...
Zi-Yi Chen, Yan Zhang, Jian-Hui Wei et al.· 1 citation· ⚡1
Dr Drift-Bench++ is introduced, a principled benchmark construction pipeline for verified executable tasks with controlled misalignment and intent shifts, along with an interaction protocol featuring finite patience, diverse simulated users, and silent interaction-conditioned shifts, which offers a foundation for evalu...
Zhe-Yuan Zhang, M. Chao, Ke Xiao et al.· arXiv.org· 4 citations
Quantitative and qualitative evaluations on multi-shot story generation demonstrate that the Complementary Retrieval-Augmented Prompting method consistently outperforms recent-frame, memory-based, and entity-level retrieval baselines in cross-shot consistency and text-controllability.
Xiang-Han Wei, Xiao-Da Yang, Zhi Wang et al.· arXiv.org· 1 citation
Robots must often continue a task after a target moves, the viewpoint shifts, or an obstacle appears, even though their earlier observations and committed actions reflect the previous scene. Many simulation robustness benchmarks fix external conditions at reset, leaving this temporal challenge underexamined. We introdu...
Yun-Bei Zhang, Zi-Jian Jin, Yuan-Zhe Liu et al.· arXiv.org· 3 citations· ⚡1
An agent can complete a task while exhibiting undesirable behavior during execution. Developers need tests for the specific behaviors encountered in deployment, beyond fixed benchmark suites. We present TraceDance, an agent system that constructs targeted benchmarks from deployment traces for user-specified undesirable...
This work proposes NOPD (Noisy Student On-Policy Self-Distillation), a simple yet effective self-distillation approach that improves VLMs without any external models or ground-truth answers, and demonstrates that NOPD is a general approach to enhance VLMs, achieving improvements across three models on 12 benchmarks.