Leandro von Werra PRO

lvwerra's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

updated a Space about 19 hours ago

Running

cowrite

✍

Create collaborative docs where AI agents suggest edits

New activity in rl-llm-wiki/rl-wiki 2 days ago

Restore in-body search + highlight (regressed by 34252e5); ?q= deep-links reuse it

#4 opened 2 days ago by

thomwolf

updated a Space 3 days ago

Running

5

Agent Manager

🖥

5

Private cloud manager for AI coding CLI sessions

updated a bucket 3 days ago

rl-llm-wiki/rl-main-bucket

318 MB

published a Space 3 days ago

Running

cowrite

✍

Create collaborative docs where AI agents suggest edits

New activity in rl-llm-wiki/knowledge-base 4 days ago

source: arxiv:2504.02495 — Inference-Time Scaling for Generalist Reward Modeling (DeepSeek-GRM/SPCT)

#660 opened 4 days ago by

lvwerra

source: arxiv:2505.02387 — RM-R1: Reward Modeling as Reasoning

#659 opened 4 days ago by

lvwerra

Add expert sub-article: generative & reasoning reward models (scalar -> verbalized judgments)

#658 opened 4 days ago by

lvwerra

source: arxiv:2506.01937 — RewardBench 2: Advancing Reward Model Evaluation

2

#656 opened 4 days ago by

lvwerra

source: arxiv:2506.16712 — ReasonGRM: Enhancing Generative Reward Models through Large Reasoning Models

2

#657 opened 4 days ago by

lvwerra

figures: sync-vs-async GPU timelines for training-systems SS3

1

#652 opened 4 days ago by

lvwerra

figures: the what-do-you-clip axis (PPO/GRPO vs DAPO vs CISPO vs GSPO) for grpo SS9

1

#651 opened 4 days ago by

lvwerra

upvoted an article 4 days ago

view article

Article

When will language models be good enough?

• 

craffel

•

4 days ago

• 4

updated a bucket 4 days ago

rl-llm-wiki/rl-the-gatherer

180 MB

New activity in rl-llm-wiki/knowledge-base 4 days ago

fix (fable audit #11): broaden tool-use-rl past single-paper + add current-status section

2

#655 opened 4 days ago by

lvwerra

deepen offline-rl: §2.1 equations + AWR->DPO derivation, §2.2 runnable check (two-tier-depth fix)

2

#654 opened 4 days ago by

lvwerra

fix (fable audit #13): reference-block ordering defects in two reward-modeling articles

2

#653 opened 4 days ago by

lvwerra

edit: strip agent-workflow residue from published prose (8 articles)

5

#642 opened 4 days ago by

lvwerra

figures: first two wiki figures (overoptimization turnover + RLHF two-KLs pipeline)

2

#648 opened 4 days ago by

lvwerra

dedup: give the Gao over-optimization law one owning article

2

#650 opened 4 days ago by

lvwerra