Boosting Omni-Modal Language Models: Staged Post-Training with Visually Debiased Evaluation

Che Liu, Lichao Ma, Xiangyu Tony Zhang, Yuxin Zhang, Haoyang Zhang, Xuerui Yang, Fei Tian

arXiv:2605.12034 · 2026-05-15 공개 · arXiv · PDF

self-distillation rlvr post-training omni-modal qwen2-5-omni benchmark-audit staged-training visually-debiased

Abstract

Omni-modal language models are intended to jointly understand audio, visual inputs, and language, but benchmark gains can be inflated when visual evidence alone is enough to answer a query. We study whether current omni-modal benchmarks separate visual shortcuts from genuine audio-visual-language evidence integration, and how post-training behaves under a visually debiased evaluation setting. We audit nine omni-modal benchmarks with visual-only probing, remove visually solvable queries, and retain full subsets when filtering is undefined or would make comparisons unstable. This yields OmniClean, a cleaned evaluation view with 8,551 retained queries from 16,968 audited queries. On OmniClean, we evaluate OmniBoost, a three-stage post-training recipe based on Qwen2.5-Omni-3B: mixed bi-modal SFT, mixed-modality RLVR, and SFT on self-distilled data. Balanced bi-modal SFT gives limited and uneven gains, RLVR provides the first broad improvement, and self-distillation reshapes the benchmark profile. After SFT on self-distilled data, the 3B model reaches performance comparable to, and in aggregate slightly above, Qwen3-Omni-30B-A3B-Instruct without using a stronger omni-modal teacher. These results show that omni-modal progress is easier to interpret when evaluation controls visual leakage, and that small omni-modal models can benefit from staged post-training with self-distilled omni-query supervision. Project page: https://cheliu-computation.github.io/omni/

한국어 요약

📋 한 줄 요약

**[옴니모달 LLM / 평가·학습]** 시각 단서만으로 풀리는 질문을 제거한 평가 셋 OmniClean과 3단계 사후학습 레시피 OmniBoost로 3B 옴니모달 모델이 30B 모델에 필적함을 보임.

🎯 핵심 기여도

💡 핵심 아이디어

"옴니모달 성능 향상"은 평가 단계의 시각 누수를 통제할 때 비로소 해석 가능해지며, 작은 모델도 self-distilled omni-query supervision을 활용한 단계적 사후학습으로 큰 모델 수준에 도달할 수 있다.

🔬 기술적 접근법

📊 주요 결과

💭 의의 및 한계

**의의**: 옴니모달 평가의 신뢰성을 회복하면서 소형 모델 사후학습의 새로운 천장을 제시. **한계**: visual-only probing이 자동화되지 않은 일부 벤치마크는 보존 처리, hazard taxonomy·언어 적용 범위는 본문 가정에 의존.

🚀 실용적 활용