AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

RL with Learnable Textual Feedback: A Bilevel Approach

Utsav Singh, Sidhaarth Sredharan Murali, Souradip Chakraborty, Amrit Singh Bedi

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 bilevel regularization (ablation)Bi-NAC · MATH-500 (Llama-3.2-1B, feedback analysis)accuracy (%): 41.80 -> 46.60Improved
RL-train the critic (higher level) while actor stays pre-trained (ablation)Hier-NFT · MATH-500 (Llama-3.2-1B, feedback analysis)accuracy (%): 35.40 -> 38.00Improved
also fine-tune the lower-level actor with the trained critic (ablation)Hier-FT · MATH-500 (Llama-3.2-1B, feedback analysis)feedback accuracy gain (%): 38.00 -> 41.80Improved
replace fixed feedback with learned feedback (ablation)SCoRe · MATH-500 (Llama-3.2-1B, feedback analysis)accuracy (%): 39.60 -> 46.60Improved
replace fixed feedback with learned feedback (ablation)SCoRe · MATH-500 (Llama-3.2-1B, feedback analysis)feedback actionability (FO): 2.1 -> 4.2Improved
use one model for both feedback and refinement (1B) instead of two models (2B) (ablation)Bi-NAC · MATH-500accuracy (%): 46.56 -> 46.84No clear effect
use one model for both feedback and refinement (1B) instead of two models (2B) (ablation)Bi-NAC · MBPPaccuracy (%): 66.73 -> 65.24No clear effect
extend refinement from turn 2 to turn 3 (ablation)Bi-NAC · MATH-500 (Qwen3-1.7B, multi-turn)accuracy (%): 70.2 -> 71.6Improved
extend refinement from turn 3 to turn 5 (ablation)Bi-NAC · MATH-500 (Qwen3-1.7B, multi-turn)accuracy (%): 71.6 -> 70.8No clear effect
use Bi-NAC feedback in refinement instead of untrained baseline at turn 2Qwen3-1.7B · MATH-500 (Qwen3-1.7B, multi-turn)accuracy (%): 65.6 -> 70.2Improved
swap LLaMA-1B critic and actor for Qwen-1.5B (ablation)Bi-NAC · MATH-500 (heterogeneous pairing)accuracy (%): 44.6 -> 48.7Mixed
Bi-NAC (2B) instead of GRPO (3B)GRPO · MATH-500 (compute trade-off)accuracy (%): 41.43 -> 46.56Improved

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