AI Paper Atlas

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

InfiMed-ORBIT: Aligning LLMs on Open-Ended Complex Tasks via Rubric-Based Incremental Training

Pengkai Wang, Pengwei Liu, Qi Zuo, Zhijie Sang, Congkai Xie, Hongxia Yang

ICML 2026 · Large language models › Instruction and fine-tuning · approach: Calibration and knowledge in LLM tuning · 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
train with ORBIT rubric-guided RL on 2k samplesQwen3-4B-Instruct · HealthBench-Hardtotal score: 7.0 -> 27.5Improved
scale ORBIT training data from 2k to 28kInfiMed-ORBIT-4B · HealthBench-Hardtotal score: 27.5 -> 33.6 -> 37.3Improved
add retrieval-augmented (human seed) rubric generation (ablation)ORBIT · HealthBench-Hardtotal score: 10.69 -> 17.89Improved
use multi-LLM synthetic contrast seeds instead of human seeds (ablation)ORBIT · HealthBench-Hard (140 steps)total score: 15.34 -> 15.60No clear effect
use a stronger training-time judge (4B self to 30B-A3B) (ablation)ORBIT · HealthBench-Hardtotal score: 13.64 -> 17.89Improved
add moderate sample pass@k filtering (0-0.75) (ablation)ORBIT · HealthBench-Hardtotal score: 20.2 -> 19.7No clear effect
add strict sample pass@k filtering (0-0.50) (ablation)ORBIT · HealthBench-Hardtotal score: 20.2 -> 14.5Hurt
add rubric-level pass@k filtering (0-0.75) (ablation)ORBIT · HealthBench-Hardtotal score: 20.2 -> 19.9No clear effect
add multi-stage restart training (ablation)ORBIT 8k · HealthBench-Hardtotal score: 25.9 -> 27.3Improved
use DeepSeek-R1 rubrics instead of GPT-5-Chat rubrics (ablation)ORBIT · HealthBench-Hard (OSS-120B)total score: 12.3 -> 20.2Improved
use rubric-guided RL instead of inference scaling K=8 to 40Qwen3-4B-Instruct · best-of-K score distributionnormalized score: ~0.65 -> beyond 0.85Improved
add ORBIT trainingQwen3-4B-Instruct · HealthBench-Hard communication qualityaxis score: -8.2Hurt

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