…Despite substantial efforts in developing video datasets and benchmark s, existing works primarily focus on perception and reasoning, without systematically evaluating memory : what models retain, how faithfully information is preserved, and how…
…Dual-Path Vision Token Routing for Multimodal Large Language Models under Visual Saturation Published on Jun 8 Submitted by Jinyang Wu on Jun 10 Authors: , Jinyang Wu Abstract Research reveals that vision…
…Generated by Qwen/Qwen2.5-Coder-32B-Instruct Scientific figures are among the most effective means of communicating complex research ideas, yet producing publication-quality illustrations remains one of the most labor…
Papers arxiv:2606.01249 Trust Region On-Policy Distillation Published on May 31 Submitted by xxr on Jun 3 Samsung Research Authors: , , , Ziheng Li , Abstract Trust Region On-Policy Distillation (TrOPD) improves…
Thanks for sharing your results and describing the background of what happened around GRPO research last week. Do you plan to test classical distillation, not just fine-tuning on reasoning traces? · Yes…
…Generated by Qwen/Qwen2.5-Coder-32B-Instruct Autonomous web navigation remains challenging for LLM agents, and the strongest generalist systems rely on proprietary reasoning models whose inference cost is prohibitive for…
…Key highlights from our research: 🏭 Grounded in Reality: 2,049 items based on Chinese national standards (GB/T) and real industrial product records (evaluated across ZH, EN, RU, and VI). ⚠️ The "Overthinking…
…I’m curious how much overhead the routing mechanism itself adds to the latency? I made a podcast on it with ResearchPod, it makes it easy to get the key concepts on…
…View arXiv page View PDF Project page GitHub 2 Add to collection Community This is a joint work with the Ohio State University and Meta Reality Labs Research. We introduce SuperMemory-VQA…
…BraveGuard mines recent research sources to identify emerging risks and attack patterns, instantiates them as executable computer-use tasks , collects agent rollouts, and derives trajectory-level supervision for guard model training. As…