AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

All Roads Lead to Likelihood: The Value of Reinforcement Learning in Fine-Tuning

Gokul Swamy, Sanjiban Choudhury, Wen Sun, Steven Wu, Drew Bagnell

ICLR 2026 · Reinforcement learning › RL from feedback · approach: RL fine-tuning of generative models · 59 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
use online DPO (RM-labeled on-policy samples) instead of offline DPOPythia on tl;dr summarization · tl;dr (gpt-4o winrate)winrate vs human referenceImproved
run online DPO on top of the offline DPO policy using an RM trained on the same preference dataoffline DPO policy · tl;dr (gpt-4o winrate)winrateImproved
train offline DPO for two epochs (DPO 2x) to match gradient steps (ablation)offline DPO · tl;dr (gpt-4o winrate)winrateNo clear effect
augment online DPO prompts with SFT-dataset prompts (about 3x data) (ablation)online DPO · tl;dr (gpt-4o winrate)winrateNo clear effect
restrict preference data to SFT-policy samples labeled by gpt-4o (narrower data) (ablation)online DPO over offline DPO · tl;dr (narrow on-policy data)winrateImproved
scale reward model size to be much smaller or larger than the policy (ablation)global RM for best-of-N · tl;dr BoNBoN winrateNo clear effect
scale up the generation policy size (ablation)best-of-N with fixed RM · tl;dr BoNBoN winrateImproved
add online DPO on two-word summarization (horizon about 1/10) (ablation)offline DPO · two-word summarization (gpt-4o labels)winrate: about +1% winrateNo clear effect
add online DPO with learned global RM when labels and eval use ROUGE-L (ablation)offline DPO · tl;dr (ROUGE-L reward)ROUGE-LNo clear effect
train offline DPO for one extra epoch (ablation)offline DPO · tl;dr (ROUGE-L reward)ROUGE-LImproved
use a global RM instead of a token-wise local RM (ablation)reward model · tl;dr validation likelihoodvalidation likelihoodImproved
add reference-policy regularization to the local RM (DPO RM vs Local) (ablation)local RM · tl;dr validation likelihoodvalidation likelihoodHurt

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