AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Jackpot: Align Actor-Policy Distribution for scalable and stable RL for LLM

Zhuoming Chen, Hongyi Liu, Yang Zhou, Haizhong Zheng, Beidi Chen

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
apply Jackpot rejection sampling for 64x rollout/train batch mismatchQwen3-4B-Base RL · AMC22 & 23 (2048/32)Mean@4: 39.15 -> 53.92 (vs off-policy)Improved
Jackpot vs TIS + Adjustment under extreme 128x mismatchQwen3-4B-Base RL · MATH-500Mean@4: 19.10 -> 80.00Improved
Jackpot vs TIS + Adjustment at 64xQwen3-4B-Base RL · GSM8KMean@4: 92.76 -> 92.24No clear effect
Jackpot vs TIS + Adjustment at 64xQwen3-4B-Base RL · AMC22 & 23Mean@4: 57.22 -> 53.92Hurt
Jackpot vs TIS + Adjustment at 64xQwen3-8B-Base RL · AMC22 & 23Mean@4: 60.54 -> 63.55Improved
add Jackpot to separate rollout and trainer models (Qwen2.5-1.5B-IT rollout, 3B-Base trainer)GRPO · MATH-500 and AMCaccuracy: up to 12% on MATH-500Improved
add reweighting on top of rejection masking under staleness (ablation)Jackpot · BF16 staleness, AMCpass@1: 49.699 -> 63.855Improved
apply rejection sampling alone under FP8 on-policy training (ablation)Jackpot masking-only · FP8 on-policy, AIME24pass@1: 23.958 -> 25.625Improved
vary truncation constant C1 in 2 to 8 (ablation)Jackpot · benchmarkspass@1No clear effect
increase top-k from 10 to 20 or 40 for Z estimate (ablation)Jackpot · AMC, MATH500pass@1: 61.446 -> 63.855 -> 61.446 (AMC)No clear effect
raise rejection threshold above 1.0 (ablation)Jackpot · AMCpass@1: 48.5 -> 45.8 (c=1.2)No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy