AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

Dissecting Post-Training: Uncovering the Complementary Roles of SFT and RL for Document Parsing

Jun-Peng Jiang, An-Yang Ji, Shiyin Lu, Guodong Zheng, Weihong Zhang, Qing-Guo Chen, Weihua Luo, Kaifu Zhang, Long Chen, De-Chuan Zhan, Han-Jia Ye

ICML 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
standard SFT on Markdown (100k synthetic pairs)Ovis2.5-2B and Qwen2.5-VL-3B · OmniDocBenchOverallNo clear effect
standard RL (GRPO, edit-distance reward)Ovis2.5-2B and Qwen2.5-VL-3B · OmniDocBenchOverallImproved
add 50k general-purpose data to SFT (ablation)SFT · OmniDocBenchOverallNo clear effect
add 10k targeted table/formula data to SFT (ablation)Markdown SFT · OmniDocBenchTEDS and CDMNo clear effect
triple SFT data from 100k to 300k (ablation)Markdown SFT · OmniDocBenchOverallNo clear effect
predict structured JSON instead of Markdown in SFT (ablation)SFT · OmniDocBenchOverallImproved
add table/formula sub-task data to JSON SFT (ablation)Ovis2.5-2B · OmniDocBenchOverall: 71.58 -> 82.43Improved
add table/formula sub-task data to JSON SFT (ablation)Qwen2.5-VL-3B · OmniDocBenchOverall: 79.41 -> 81.09Improved
use decomposed element-wise reward instead of global reward (ablation)RL, Ovis2.5-2B · OmniDocBenchOverall: 66.59 -> 71.62Improved
use decomposed reward instead of global reward (ablation)RL, Ovis2.5-2B · OmniDocBenchTable TEDS: 71.3 -> 60.97Hurt
use decomposed reward instead of global reward (ablation)RL, Qwen2.5-VL-3B · OmniDocBenchOverall: 84.11 -> 84.29No clear effect
add table/formula data to RL with decomposed reward (ablation)RL, Ovis2.5-2B · OmniDocBenchOverall: 71.62 -> 79.00Improved

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