Skip to content

feat(daily): 2026-05-04 digest#29

Open
yayajjiang wants to merge 1 commit into
mainfrom
digest/daily-2026-05-04
Open

feat(daily): 2026-05-04 digest#29
yayajjiang wants to merge 1 commit into
mainfrom
digest/daily-2026-05-04

Conversation

@yayajjiang
Copy link
Copy Markdown
Owner

Papers added (2026-05-04)

  • ResRL (2605.00380) — SVD负样本投影解耦正负样本语义,12项数学/代码/智能体基准全面超越GRPO/NSR。⭐ Pick
  • AGoQ (2605.00539) — 逐层量化激活与梯度,训练内存降低52%,速度比Megatron-LM提升1.34倍。
  • Reward Models Are Secretly Value Functions / TCRM (2604.22981) — 奖励模型本质是价值函数,TCRM正则化统一token级评分与RL条件期望。

https://claude.ai/code/session_01QnCCbRwF7Jd8yNfek2r5zy


Generated by Claude Code

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants