AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Breaking Barriers: Do Reinforcement Fine-tuning Gains Transfer To Unseen Domains?

Chuxuan Hu, Yuxuan Zhu, Antony Kellermann, Caleb Biddulph, Suppakit Waiwitlikhit, Jason Benn, Daniel Kang

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data

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 RPT on in-domain versus out-of-domain tasks18 open RPT models · Observational studypass@1 gain: 2.87 in-domain vs -3.19 out-of-domainMixed
add single-domain RPT on math, evaluated out-of-domainDeepSeek-R1-Distill-Qwen-1.5B + GRPO · OOD taskspass@1 gain: 3.13 in-domain -> -1.81 out-of-domainHurt
train with math RPT, evaluated on codemath-RPT models · Observational study, code taskspass@1: 4.77% improvement on codeImproved
train with code RPT, evaluated on mathcode-RPT models · Observational study, math taskspass@1: 9.49 on code, 15.44 on mathImproved
train with math or code RPT, evaluated on knowledge tasksmath/code-RPT models · Observational study, knowledge taskspass@1: -0.27 vs 11.08 on math and 5.82 on codeNo clear effect
train with knowledge RPT, evaluated on math and codeknowledge-RPT models · Observational studypass@1: 21.40 on math, 12.16 on codeImproved
train with knowledge RPT, evaluated on other knowledge subdomainsFino1-8B · Observational studyTabFact accuracy: 64.18% -> 48.39%Hurt
switch RL algorithm from GRPO to DAPO (ablation)DeepSeek-R1-Distill-Qwen-1.5B math RPT · Table 3ID-OOD gain gap: 4.94 -> 5.23No clear effect
switch base model from DeepSeek-R1-Distill-Qwen-1.5B to Llama-3.2-3B-Instruct (ablation)math RPT with GRPO · Table 3ID-OOD gain gap: 4.94 -> 5.06No clear effect
increase RPT training from 0.5 to 2 epochs (ablation)Math-RPT · Figure 5OOD pass@1 gain: -0.63 -> -1.47 OOD; 3.19 -> 4.72 IDHurt
scale up model size within a family (ablation)open RPT models · Table 4ID-OOD gain gap: 1.58 average gap difference (Large - Small)No clear effect
train on mixed-domain data instead of the best single domain (ablation)Mixed-RPT · Table 5pass@1 gain on knowledge and code: knowledge -3.98 -> -2.17; code -0.02 -> 1.91; math -0.15 -> -3.62Mixed

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