…Across two deep research benchmarks, DuMate-DeepResearch establishes new state-of-the-art results: the best overall score (58.03%) on DeepResearch Bench , and the best overall score (61.95%) on DeepResearch…
…Measuring Reward Hacking in Long-Horizon Coding Agents (2026) Reward Hacking Benchmark: Measuring Exploits in LLM Agents with Tool Use (2026) Do Synthetic Trajectories Reflect Real Reward Hacking? A Systematic Study on…
…Wangcheng Tao , , Abstract Self-Evolving Prompt Optimization (SePO) enhances agent performance by jointly optimizing both task and prompt agent system prompts through evolutionary search, demonstrating superior accuracy across diverse benchmarks. Generated by…
…We recently launched the Ugandan Cultural Context Benchmark (UCCB) on Hugging Face ( https://huggingface.co/datasets/CraneAILabs/UCCB ), which represents a significant milestone for African AI evaluation. As the first comprehensive benchmark…
…We demonstrate that MACU consistently improves over strong single-agent baselines by 3.4-25.5% on desktop (OSWorld) and web navigation (Online-Mind2Web, WebTailBench, Odysseys) benchmarks, exhibits more favorable test-time…
…A Benchmark for Executable Causal Mechanism Induction from Interventions (2026) NoisyCausal: A Benchmark for Evaluating Causal Reasoning Under Structured Noise (2026) Push Your Agent: Measuring and Enforcing Quantitative Goal Persistence in Long…
…CoE benchmark (51K multi-hop QA with bbox annotations), and significantly outperforms text-based baselines on SlideVQA where layout matters. We hope CoE pushes RAG from information finding toward causal, verifiable visual…
…AI-generated summary Speculative decoding speeds up autoregressive generation in Large Language Models (LLMs) through a two-step procedure, where a lightweight draft model proposes tokens which the target model then verifies…
…Asymmetric Group Policy Optimization for Verifiable Reasoning and Search Ads Relevance at JD (2026) CoVerRL: Breaking the Consensus Trap in Label-Free Reasoning via Generator-Verifier Co-Evolution (2026) Please give a…
…We therefore evaluate attacks by Attack Success Rate , which requires the injected payload to execute and the user's task to still pass its verifier in the same trial. Prior skill-poisoning…