… View arXiv page View PDF Add to collection Community seeing how much the gains hinge on draft initialization and keeping drafts short, it looks like the method pays off most when the speculative drafts stay close to the current rollout distribution. my question is: how does speculative decoding beh… …
… We show that practical asynchronous pipelines with delayed updates and partial rollouts often lose the required historical training-side logits, or old logits. …
… AI-generated summary Reinforcement learning with verifiable rewards RLVR for Large Reasoning Models hinges on baseline estimation for variance reduction , but existing approaches pay a heavy price: PPO requires a policy-model scale critic , while GRPO needs multiple rollouts per prompt to keep its … …
… View arXiv page View PDF Project page Add to collection Community ClawGym: full-lifecycle framework for Claw-style personal agents 1️⃣ Data : ClawGym-SynData — 13.5K filtered tasks, persona-driven intents × skill-grounded ops, with mock workspaces & hybrid verification 2️⃣ Training : ClawGym-Agents… …
… A dual-forward-pass strategy uses teacher forcing to evaluate corrupted views against clean-image trajectories, avoiding new rollouts on degraded inputs. …
… Our analyses further show that direct set-level diversity rewards remain effective with small groups and help preserve broader generation-distribution coverage during post-training. The code is available at https://github.com/IDEA-XL/S GRPO . …
… AI-generated summary On-policy distillation offers dense, per-token supervision for training reasoning models; however, it remains unclear under which conditions this signal is beneficial and under which it is detrimental. …
… View arXiv page View PDF GitHub 10 Add to collection Community Overview we propose A²TGPO A gentic T urn- G roup P olicy O ptimization with A daptive Turn-level Clipping , which retains IG as the intrinsic signal but re-designs how it is normalized, accumulated, and consumed: i turn-group normaliza… …
… Interesting breakdown of this paper on arXivLens: https://arxivlens.com/PaperView/Details/tmas-scaling-test-time-compute-via-multi-agent-synergy-682-9aa2c46a Covers the executive summary, detailed methodology, and practical applications. …
… AI-generated summary Large reasoning models , such as OpenAI o1 and DeepSeek-R1, tend to become increasingly verbose as their reasoning capabilities improve. …