ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decoding

Yiran Guan; Sifan Tu; Dingkang Liang; Linghao Zhu; Jianzhong Ju; Zhenbo Luo; Jian Luan; Yuliang Liu

cs.CVFeb 26, 2026

ThinkOmni: Lifting Textual Reasoning to Omni-modal Scenarios via Guidance Decoding

Authors: Yiran Guan, Sifan Tu, Dingkang Liang, Linghao Zhu, Jianzhong Ju, Zhenbo Luo, Jian Luan, Yuliang Liu,

Organizations: Huazhong University of Science and Technology · MiLM Plus, Xiaomi Inc.

Abstract

Omni-modal reasoning is essential for intelligent systems to understand and draw inferences from diverse data sources. While existing omni-modal large language models (OLLM) excel at perceiving diverse modalities, they lack the complex reasoning abilities of recent large reasoning models (LRM). However, enhancing the reasoning ability of OLLMs through additional training presents significant challenges, including the need for high-quality data, task-specific adaptation, and substantial computational costs. To address these limitations, we propose ThinkOmni, a training-free and data-free framework that lifts textual reasoning to omni-modal scenarios. ThinkOmni introduces two key components: 1) LRM-as-a-Guide, which leverages off-the-shelf LRMs to guide the OLLM decoding process; 2) Stepwise Contrastive Scaling, which adaptively balances perception and reasoning signals without manual hyperparameter tuning. Experiments on six multi-modal reasoning benchmarks demonstrate that ThinkOmni consistently delivers performance improvements, with main results achieving 70.2 on MathVista and 75.5 on MMAU. Overall, ThinkOmni offers a flexible and generalizable solution for omni-modal reasoning and provides new insights into the generalization and application of reasoning capabilities. Code is publicly available at https://github.com/1ranGuan/thinkomni

Figures & tables

Figure 1: Comparison between existing methods and ThinkOmni . We integrate an OLLM with an LRM via guidance decoding, enabling advanced reasoning abilities with omni-modal input.

Figure 3: Guidance decoding methods. “Guid.” denotes the guiding model, and “Amat.” denotes the amateur model.

Figure 4: Overview of ThinkOmni . The framework begins by separating input modalities of the OLLM and introducing the LRM as a guiding model. Stepwise Contrastive Scaling dynamically adjusts guidance parameters based on real-time prediction analysis, enabling adaptive and effective decoding across diverse tasks.

Figure 5: Case studies from (a) MMAU and (b) MathVista. ( Sakshi et al., 2025 ; Lu et al., 2024 ) Tasks require different levels of LRM involvement. Using a fixed $\alpha$ limits the ability of the model to optimally adapt to task-specific needs, highlighting the need for a more flexible approach.

Figure 6: Detailed process of ThinkOmni . (a) The OLLM handles multi-modal inputs, while the LRM focuses on textual reasoning. By mixing their logits, the system effectively integrates perception and reasoning during token generation. (b) Each decoding step balances perception and reasoning dynamically by comparing logit distributions under different conditions and models.

Model	MathVista test-mini	MathVision test	MathVerse test-mini	MMAU (v05.15.25) test-mini	DailyOmni test	OmniBench test
Close-Sourse Models
GPT-4o	63.8	30.4	50.8	62.5	56.5	-
Gemini-2.0-Flash	73.1	41.3	59.3	70.5	67.8	-
Open-Sourse Omni Models
Baichuan-Omni-1.5	63.6	-	-	66.2	50.0	42.9
Ola	68.4	-	-	70.3	50.71	-

Table 1: Model performance on several omni-modal reasoning benchmarks. Here, DeepSeek refers to DeepSeek-R1-Distill-Qwen-7B ( Guo et al., 2025 ) , and Qwen3 denotes Qwen3-8B ( Yang et al., 2025a ) . The numbers in parentheses indicate the performance changes compared to the base OLLMs Qwen2.5-Omni-3B / 7B ( Xu et al., 2025 ) and Omni-R1 ( Zhong et al., 2025 ) . Models marked with ‘ ∗ ’ are evaluated using our own evaluation scripts.

Method	MathVista test-mini	MMAU (v05.15.25) test-mini	OmniBench test
Base Model	66.8	71.5	42.1
Average Logits Fusion	55.0( -11.8 )	55.7( -15.8 )	36.1( -6.0 )
Caption-then-Answer	61.0( -5.8 )	59.7( -11.8 )	32.3( -9.8 )
VCD	66.5( -0.3 )	72.2( +0.7 )	43.1( +1.0 )
ThinkOmni (Ours)	68.8( +2.0 )	73.8( +2.3 )	43.2( +1.1 )

Table 2: Comparison with several training-free methods. All are built upon Qwen2.5-Omni-7B.

Figure 7: Ablation on guidance weight. Left: Performance varies with the constant guidance weight

\alpha

. Each task’s optimal

\alpha

range differs, with

\alpha=0

as the OLLM baseline. Right: ThinkOmni uses adaptive dynamic weights, and the dynamic

\alpha^{r}

shows a similar distribution shift, indicating that stepwise contrastive scaling can flexibly adapt to different task requirements.

Figure 8: LRM-as-a-Guide performance scaling. We replace the Guide LRM on several benchmarks to study the impact of different LRM sizes and different LRM series on the performance of our method. The baseline refers to the performance of the Qwen2.5-Omni-7B

Figure 9: An OmniBench case study. This case study visualizes the reasoning process of ThinkOmni ‑Qwen2.5-Omni‑7B, highlighting the stepwise contrastive scaling coefficient $\alpha^{r}$ .

Figure 10: Failure case study. (a) ThinkOmni makes a reasoning error due to conflicting information between the visible beaker markings (400ml) and the label (600ml). (b) ThinkOmni fails because it cannot accurately detect the drum kit’s true start time in the audio. 2 2 2 For the reader’s understanding, the colored markings in Fig. 2 are visual aids added afterward and were not provided to the model.

Guidence Decoding Method	Model Size	Prefill Time $\downarrow$	Generate Time $\downarrow$
None (Baseline)	7B	0.138s	0.025s
VCD ( Leng et al., 2024 )	7B	0.262s ( $1.89\times$ )	0.050s ( $2.00\times$ )
ProxyTuning ( Liu et al., 2024 )	7B + 3B + 3B	0.406s ( $2.94\times$ )	0.086s ( $3.44\times$ )
ThinkOmni (Ours)	7B +7B	0.191s ( $1.38\times$ )	0.072s ( $2.88\times$ )
ThinkOmni (Ours)	7B +1.5B	0.191s ( $1.38\times$ )	0.069s ( $2.76\times$ )

Table 3: Generation latency comparison. Prefill Time: first token generation latency; Generate Time: response generation with KV cache. Results are averaged over 100 samples from OmniBench.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Multimodal Reasoning

Multimodal reasoning studies how AI systems combine text with images, audio, video, and other signals to solve evidence-dependent tasks.

See all 506 papers

Text-Centric Post-Training for Omni-Modal Reasoning
Oct 2, 2026Ziyang Cheng, Yuhao Wang, Hongcheng Liu +5Multimodal Reasoning Audio-Visual Understanding
Shanghai Jiao Tong University
OmniThoughtVis: A Scalable Distillation Pipeline for Deployable Multimodal Reasoning Models
May 12, 2026Yuanhao Yue, Chengyu Wang, Yuanjie Lyu +2Multimodal CoT Reasoning Multimodal Reasoning
Alibaba Group, Hangzhou, China
Cross-Modal Coreference Alignment: Enabling Reliable Information Transfer in Omni-LLMs
Apr 7, 2026Hongcheng Liu, Yuhao Wang, Zhe Chen +5Cross-Modal Alignment Multimodal Large Language Models
Shanghai Jiao Tong University

Oct 2, 2026J/K move · Enter open · S save

Text-Centric Post-Training for Omni-Modal Reasoning

Ziyang Cheng, Yuhao Wang, Hongcheng Liu, Qimin Wu+4

Shanghai Jiao Tong University

Multimodal Reasoning Audio-Visual Understanding RL for Language Model Reasoning

Open paper AskPDF arXiv

Improving joint audio-visual reasoning in Omni Large Language Models typically incurs substantial data construction and training costs. Our diagnostics reveal multi-hop reasoning difficulties despite correct answers to all corresponding single-hop questions and suggest partial decoupling in the local optimization of perception and reasoning objectives. This motivates post-training with different emphases on these capabilities. Text-only reasoning training yields gains across data sources, model scales, and families. With the best-performing text-only configuration, supervised fine-tuning followed by reinforcement learning (RL) raises Qwen2.5-Omni-7B's geometric mean of nine reasoning scores by 25.83% over the base model, outperforming the complete native audio-visual route with 56.6% fewer GPU-hours. Training on data synthesized entirely by a text-only LLM raises this geometric mean by 21.01% without audio-visual data in construction or training. However, text-only training degrades perception. We therefore propose a text-centric post-training paradigm: text-only training provides the main reasoning optimization, and reduced-data native audio-visual RL then refines perception. Refinement uses about 90% fewer input tokens than full-data audio-visual RL, restores perception above the base level, and retains 93.5% of the best-performing text-only pipeline's reasoning gain.

Abstract

Multimodal Reasoning

Explore similar work

Text-Centric Post-Training for Omni-Modal Reasoning

OmniThoughtVis: A Scalable Distillation Pipeline for Deployable Multimodal Reasoning Models

Cross-Modal Coreference Alignment: Enabling Reliable Information Transfer in Omni-LLMs