DeAR: Decentralized Agentic Reasoning via Capability Grounding and Collaborative Thought Navigation

Xing Wei; Changmeng Zheng; XiaoYong Wei; Xiufen Ye; Qing Li

cs.AIAug 18, 2026

DeAR: Decentralized Agentic Reasoning via Capability Grounding and Collaborative Thought Navigation

Authors: Xing Wei, Changmeng Zheng, XiaoYong Wei, Xiufen Ye, Qing Li

Organizations: College of Intelligent Systems Science and Engineering, Harbin Engineering University · Department of Computing, The Hong Kong Polytechnic University

Abstract

Existing agentic reasoning systems typically rely on centralized protocols. This design introduces routing bottlenecks and static role allocations that often fail when handling complex multimodal queries. We propose DeAR (Decentralized Agentic Reasoning), a framework that shifts from central control to autonomous peer-to-peer collaboration. DeAR is built on three mechanisms: (1) decentralized capability grounding for query-dependent agent specialization, (2) thought map navigation for targeted peer interactions, and (3) topology update for adaptive error correction. Evaluations across 9 diverse multimodal reasoning and text-based QA benchmarks indicate that DeAR consistently outperforms recent baseline methods, validating that decentralized and adaptive collaboration among agents enhances accuracy in knowledge-intensive reasoning tasks. The source code will be available at https://open_upon_acceptance.

Figures & tables

Figure 1: Framework comparison on a multimodal query ("How many people in the image were born after World War II?"). (a) Single agent fails due to restricted knowledge domains. (b) Centralized framework fails as the judge bottleneck propagates incorrect inputs. (c) DeAR (Ours) succeeds by grounding capabilities and navigating an adaptive thought map to collaboratively verify facts and yield the correct answer.

Figure 2: An overview of the De centralized A gentic R easoning framework.

Figure 3: Thought Map Navigation Construction among Agents

Method	MMMU	MathVista						ChartQA	ScienceQA
Method	MMMU	FQA	GPS	MWP	TQA	VQA	ALL	ChartQA	NAT	SOC	LAN	ALL
Baseline Models (Single Agent)
LLaMA-2-13B ( Liu et al. 2023 )	-	26.8	29.3	16.1	32.3	26.3	26.10	-	44.1	41.2	43.9	43.08
InternVL2.5-26B ( Chen et al. 2024b )	51.80	38.0	35.0	30.0	40.0	37.0	36.03	-	-	-	-	-
MiniCPM-V-2.6 ( Yao et al. 2024 )	45.11	51.7	27.4	39.8	42.5	34.7	39.89	-	59.5	57.0	59.6	58.70
Qwen2-VL-72B ( Wang et al. 2024a )	46.21	55.9	34.7	29.7	58.8	42.4	41.34	88.30	-	-	-	-

Table 1: Zero-shot evaluation on four multimodal benchmarks (fine-grained metrics reported for MathVista and ScienceQA). ’-’ denotes unreported data. Agent counts: Cache-of-Thought (2); MUG, C2R (3); MAD-Vote, Corvid, Insight-V, AutoGen, AgentVerse, DyLAN, ReConcile, and DeAR (4). Best results in bold .

Single Agent Framework
Method	NQ				TriviaQA				PopQA				2Wiki				HotpotQA
	EM	$\Delta$ EM	F1	$\Delta$ F1	EM	$\Delta$ EM	F1	$\Delta$ F1	EM	$\Delta$ EM	F1	$\Delta$ F1	EM	$\Delta$ EM	F1	$\Delta$ F1	EM	$\Delta$ EM	F1	$\Delta$ F1
Vanilla Gen	17.40	+29.79	26.27	+27.86	56.60	+4.87	65.50	+7.54	19.20	+20.54	23.33	+23.87	8.60	+42.53	16.50	+44.74	16.80	+31.51	23.88	+29.86
Multi-Agent Frameworks
MAD ( Du et al. 2023 ) (ICML 2024)	21.80	+25.39	33.11	+21.02	56.40	+5.07	66.39	+6.65	21.40	+18.34	28.28	+18.92	18.20	+32.93	25.13	+36.11	23.00	+25.31	32.79	+20.95
IRCoT ( Trivedi et al. 2023 ) (ACL 2023)	28.60	+18.59	37.36	+16.77	47.20	+14.27	54.56	+18.48	27.00	+12.74	33.02	+14.18	32.80	+18.33	31.19	+30.05	25.20	+23.11	34.40	+19.34

Table 2: Overall evaluation results of the proposed DeAR framework and other baselines on five text-only QA benchmarks.

\Delta

EM and

\Delta

F1 denote the performance gap between DeAR and each baseline (

\Delta=\text{DeAR score}-\text{baseline score}

). Darker gray cells mark the best performance, and lighter gray cells mark the second-best performance among all methods.

Figure 4: Analysis of the topology update’s decay factor

\alpha

Figure 5: Case Study of the Reasoning with Centralized and Decentralized Multi-Agent Collaboration on Q&A.

Number of Agents Activated	Frequency (%)
1 Agent	14.3
2 Agents	21.8
3 Agents	26.4
4 Agents	37.5

Table 3: Frequency of activation for varying numbers of agents on the MathVista dataset.

Multi-Agent Reasoning

Multi-Agent Reasoning studies how multiple AI agents share evidence, model one another, and coordinate decisions on complex tasks.

See all 233 papers

Adaptive Latent Agentic Reasoning
Jun 1, 2026Dongwon Jung, Peng Shi, Yi Zhang +2Agentic Reasoning LLM Agents
University of California, Davis · University of Waterloo · Greenshoe, Inc.
RUMAD: Reinforcement-Unifying Multi-Agent Debate
Feb 27, 2026Chao Wang, Han Lin, Huaze Tang +2Communication Topology Optimization Multi-Agent Debate
Tsinghua University, Shenzhen, China · Zhejiang University, Hangzhou, China
DeALOG: Decentralized Multi-Agents Log-Mediated Reasoning Framework
Feb 1, 2026Abhijit Chakraborty, Ashish Raj Shekhar, Shiven Agarwal +1Multi-Agent Collaboration Multimodal QA
Arizona State University

Jun 1, 2026J/K move · Enter open · S save

Adaptive Latent Agentic Reasoning

Dongwon Jung, Peng Shi, Yi Zhang, Junshan Zhang+1

University of California, Davis · University of Waterloo · Greenshoe, Inc.

Agentic Reasoning LLM Agents LLM Agent Training Efficient Language Model Inference Latent CoT Reasoning

Open paper AskPDF arXiv

Large reasoning models improve performance by generating extended chain-of-thought (CoT) reasoning, but this behavior becomes inefficient when applied to LLM agents. Current LLM agents often generate verbose textual reasoning at every decision step and allocate reasoning effort nearly uniformly across turns, leading to substantial inefficiency in multi-turn agentic trajectories. We propose Adaptive Latent Agentic Reasoning (ALAR), a dual-mode framework that uses compact latent reasoning for routine turns and selectively escalates to explicit chain-of-thought when deeper deliberation is needed. ALAR learns latent reasoning by using the agent's actions as supervision anchors and is further optimized to use latent reasoning when it is sufficient for task success and reserve explicit CoT for harder decisions. Experiments on agentic search and tool-use benchmarks show that ALAR maintains comparable or better task accuracy while substantially reducing generated tokens by up to 43.6% in search and 84.6% in tool use. These results demonstrate that ALAR improves the accuracy-efficiency trade-off of LLM agents by reducing unnecessary textual reasoning while preserving explicit deliberation for harder decision steps.

Abstract

Multi-Agent Reasoning

Explore similar work

Adaptive Latent Agentic Reasoning

RUMAD: Reinforcement-Unifying Multi-Agent Debate

DeALOG: Decentralized Multi-Agents Log-Mediated Reasoning Framework