AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Beyond Token-level Supervision: Unlocking the Potential of Decoding-based Regression via Reinforcement Learning

Ming Chen, Sheng Tang, Rong-Xi Tan, Ziniu Li, Jiacheng Chen, Ke Xue, Chao Qian

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 0 citations (OpenAlex)

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
add sequence-level RL (ReMax) finetuning after CEdecoding-based regressor · TALENT (100 regression tasks, median aggregation)RMSE: 0.5484 -> 0.5190Improved
add sequence-level RL (GRPO) finetuning after CEdecoding-based regressor · TALENT (100 regression tasks, median aggregation)R2: 0.6124 -> 0.6248Improved
add token-level NTL-WAS finetuningdecoding-based base model · TALENT (100 regression tasks, median aggregation)RMSE: 0.5484 -> 0.5478No clear effect
add DIST2 token-level lossdecoding-based base model · TALENT (100 regression tasks, mean aggregation)R2: 0.6368 -> 0.4678Hurt
replace pointwise head with GenRe2-ReMax decoderregression head · TALENT (100 regression tasks, median aggregation)R2: 0.5708 -> 0.6459Improved
add reward standardization (GRPO) (ablation)GenRe2-GRPO · TALENT (100 tasks, digit base 10)R2: 0.6089 -> 0.5872Hurt
add importance-sampling clipping (ablation)GenRe2-GRPO · TALENT (100 tasks, digit base 10)R2: 0.5854 -> 0.5872No clear effect
use greedy baseline instead of mean baseline (ablation)GenRe2-GRPO · TALENT (100 tasks, digit base 10)R2: 0.5872 -> 0.5876No clear effect
increase normalized-tokenization digit base to 10 (ablation)GenRe2-GRPO · TALENT (100 tasks)R2Hurt
add GenRe2-ReMax finetuningRLM code regression · APPS LeetcodeRMSE: 0.493 -> 0.474Improved
finetune with CE on datasetRLM code regression · Triton Kernel LatencyRMSE: 1.095 -> 16.37Hurt
add GenRe2-ReMax finetuningRLM code regression · Triton Kernel LatencyRank correlation: 0.536 -> 0.598Improved

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