LongCat-DeepResearch Technical Report

Meituan LongCat Team; He Zhu; Yue Xu; Wanli Wu; Haolin Ren; Yuxin Bian; Jiarui Zhao; Rongzhi Zhang

cs.AISep 28, 2026

LongCat-DeepResearch Technical Report

Authors: Meituan LongCat Team, He Zhu, Yue Xu, Wanli Wu, Haolin Ren, Yuxin Bian, Jiarui Zhao, Rongzhi Zhang, +23 more

Organizations: Meituan LongCat Team

Abstract

We present LongCat-DeepResearch, a deep research system that combines an enhanced LongCat model with a multi-agent workflow for producing comprehensive, evidence-grounded reports. The workflow separates global planning from detailed investigation and coordinates revision at the section level. Multiple planning agents first explore external sources and refine an actionable research plan, termed ResearchSpec. Research agents then investigate and draft their assigned sections in parallel, gathering additional evidence in separate contexts as their analyses develop. Once the sections are assembled, global review guides targeted local revisions, reducing reliance on repeated full-report rewriting. This workflow also supports the construction of research tasks and trajectories for the mid-training and post-training of LongCat's general-purpose models. LongCat-DeepResearch achieves 55.25 on DeepResearchBench, 51.35 on DeepResearchBench II, and 79.83 on ResearchRubrics. On an in-house benchmark, it scores 76.04, ranking second among four compared systems. Development-set analyses show benefits from combining planning perspectives, while further planning refinement has mixed effects. Additional editing improves average automatic readability preference across two benchmarks, with different trends on each.

Figures & tables

Figure 1: Overall scores on DeepResearchBench, DeepResearchBench II, and ResearchRubrics (higher is better). Green bars highlight LongCat-DeepResearch; gray bars show the three comparison systems. Evaluation settings and scored coverage are detailed in Section 4 and Appendix B .

DeepResearchBench
Metric	Gemini- DeepResearch	ChatGPT- DeepResearch	Claude- DeepResearch	LongCat- DeepResearch
Comprehensiveness	49.47	55.41	53.71	56.21
Insight / Depth	50.73	55.82	53.87	56.48
Instruction following	50.88	54.99	53.59	55.18
Readability	49.28	51.51	51.35	49.96
Overall	50.21	54.95	53.43	55.25

Table 1: Public and in-house benchmark results with benchmark-specific dimensions; higher is better. In-house scores use automatic evaluation ( Section 4.5 ); Overall is the weighted total. The descriptive average is the unweighted mean of the four Overall scores, which use different scoring scales and protocols. The Gemini, ChatGPT, and Claude results are obtained through their official clients using each provider’s own Deep Research system, with Gemini 3.7 Flash, GPT-5.6 Sol (xhigh), and Claude Opus 5 (xhigh) selected, respectively. See Appendix B for evaluation details.

Configuration	DRB-II	ResearchRubrics	Avg.
Full	48.68	79.14	63.91
One Writer $\rightarrow$ Reviser	44.65	74.46	59.56
One whole-report Researcher	45.64	78.49	62.07
No Editor	48.60	78.00	63.30

Table 2: LongCat component scores; higher is better. DRB-II denotes DeepResearchBench II, and

\rightarrow

denotes stage order. Avg. is the unweighted mean of the two displayed benchmark scores. Bold marks the highest mean per column. Detailed configurations are given in Section B.2 .

Stage	ResearchSpec	ResearchRubrics
	Coverage (%)	Explicit	Implicit	Total
W0	53.76	84.34	82.21	81.67
J0	60.47	89.09	87.11	85.13
R1	63.14	91.59	82.76	84.34
R2	62.48	90.91	87.99	84.89

Table 3: LongCat planning and report scores on a fixed ResearchRubrics development subset. W0 is the single-Writer plan, J0 the Judge-integrated plan, and R1/R2 one/two Critic/Reviser cycles. ResearchSpec coverage evaluates the plan before report generation ( Section B.4 ). RR Explicit and Implicit score the corresponding report-rubric categories; Total uses the full rubric. Bold marks the best stage for each metric. The protocol is given in Section B.6 .

Configuration	Native report score			Readability
	DRB-II	RR	Avg.	Avg. preference
A0: No Editor	55.87	85.91	70.89	—
A1: Original Editor (1 round)	56.04	85.66	70.85	50.00
B1: Enhanced (1 round)	55.15	85.18	70.17	50.42
B2: Enhanced (2 rounds)	55.68	86.87	71.28	53.33
B3: Enhanced (3 rounds)	58.38	86.00	72.19	53.96

Table 4: LongCat Editor scaling. DRB-II (DeepResearchBench II) and RR (ResearchRubrics) are native report scores on the development subsets; their Avg. is the unweighted mean of the two displayed scores. Readability is the mean automatic preference score across the development subsets against the same A0 draft; it is a transformed preference margin, not a win rate. 50 denotes parity ( Section B.5 ). A dash marks A0 as the ungraded reference. Bold marks the highest mean per column, without implying statistical significance.

Model	Harness	DRB-I	DRB-II	RR	Avg.
LongCat (previous release)	ReAct / direct report	47.65	33.13	60.33	47.04
LongCat (previous release)	Current harness	53.58	48.68	71.90	58.05
LongCat (current)	Current harness	55.25	51.35	79.83	62.14

Table 5: Report quality across model and harness configurations. Previous release refers to LongCat-2.0. All three configurations are evaluated on the full benchmarks. The last row is LongCat-DeepResearch and reproduces the main results from Table 1 . Avg. is the unweighted mean of the three displayed benchmark scores. Bold marks the highest score per column.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Official leaderboard snapshot (September 14, 2026)
System	Comp.	Insight	Inst.	Read.	Overall
ATH-Voicepica-DeepResearch	56.37	56.77	55.92	53.27	55.99
cellcog-max	56.34	57.08	55.30	51.94	55.78
SPIA DeepResearch	55.26	55.99	55.12	52.70	55.12
WhaleCloud-DocChain_0612	55.14	55.33	54.85	52.48	54.78
deep-dog-2	54.68	55.32	54.26	51.05	54.32

Appendix

Table 6: DeepResearchBench RACE results under GPT-5.5 (medium) evaluation. Comp., Inst. and Read. denote comprehensiveness, instruction following and readability. Official rows preserve their published identities; the lower group uses the system snapshots in Section 4.2 .

Official leaderboard snapshot (September 14, 2026)
System	Info.	Analysis	Pres.	Overall
GPT-o3 Deep Research	38.05	46.37	83.49	43.00
Gemini 3 Pro Deep Research	36.58	43.91	84.94	41.40
Gemini 2.5 Pro Deep Research	34.28	49.11	77.98	39.98
Doubao Deep Research	32.58	44.12	71.53	37.36
Perplexity Research	29.16	36.26	73.41	33.65

Appendix

Table 7: DeepResearchBench II results under GPT-5.5 (medium) evaluation. Info. and Pres. denote information recall and presentation. The upper group is the official text-only snapshot; the lower group reproduces our main results.

Benchmark	Difference	Mean change	Descriptive 95% interval
DRB-II	A1 $-$ A0	$+0.17$	$[-0.71,\ +1.11]$
DRB-II	B3 $-$ A0	$+2.51$	$[+0.63,\ +4.78]$
DRB-II	B1 $-$ A1	$-0.89$	$[-2.51,\ +0.56]$
DRB-II	B2 $-$ B1	$+0.52$	$[-0.57,\ +1.70]$
DRB-II	B3 $-$ B2	$+2.70$	$[+0.51,\ +5.24]$
ResearchRubrics	A1 $-$ A0	$-0.25$	$[-2.08,\ +1.89]$

Appendix

Table 8: LongCat paired native-score changes on fixed development subsets. A0 is the unedited draft; A1 applies the original Editor to A0 once. B1–B3 are successive enhanced-Editor rounds starting from A0, independently of A1. Each difference subtracts the second condition from the first; positive changes favor the first condition. DRB-II denotes DeepResearchBench II.

Explore similar work

DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning
Jun 5, 2026Lingyong Yan, Can Xu, Yukun Zhao +13Deep Research Model-Based Multi-Agent Systems
DuMate Team, Baidu AI Cloud
S1-DeepResearch: Beyond Search, Toward Real-World Long-Horizon Research Agents
Jun 13, 2026Yao Dong, Xinglin Xiao, Liwei Dong +5Deep Research Long-Horizon Agents
XScience Lab · Wenge AI
DeepResearch Agent System
Jul 30, 2026Yong Huang, Yulu Huang, for the team CollaborationDeep Research Deepseek

Jun 5, 2026, cs.AIJ/K move · Enter open · S save

DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning

Lingyong Yan, Can Xu, Yukun Zhao, Wenxuan Li+12

DuMate Team, Baidu AI Cloud

Deep Research Model-Based Multi-Agent Systems Tool-Grounded Reasoning Recursive Models

Open paper AskPDF arXiv

Deep Research (DR) has emerged as a new agentic paradigm to tackle complex, open-ended research tasks, demanding systems that can iteratively frame problems, acquire evidence, verify sources, and synthesize long-form reports. In practice, however, current DR systems are constrained by four interrelated limitations: long-horizon planning over an underspecified scope, the bottleneck of decomposing and scheduling such tasks within a single agent, hallucination risk in long-form synthesis, and limited process auditability. This technical report presents DuMate-DeepResearch, a multi-agent DR framework built on the Qianfan Agent Foundry. The framework decouples the Agent Core, which handles task understanding, planning, and scheduling, from an extensible Tool Ecosystem for retrieval, evidence acquisition, and report rendering, making every intermediate decision and tool invocation explicitly traceable. Building on this infrastructure, DuMate-DeepResearch further introduces three mechanisms: (i) a graph-based dynamic planning strategy expands the research roadmap coarse-to-fine and continuously revises it through reflection, re-planning, backtracking, and parallel branching; (ii) a recursive two-level execution design delegates each complex search sub-task to an inner Search Agent that runs its own planning loop, isolating noisy retrieval and stabilizing long-horizon execution; (iii) a rubric-based test-time optimization mechanism dynamically generates task-specific quality criteria and uses them as live reasoning scaffolds for evidence-grounded synthesis and adaptive stopping. Across two deep research benchmarks, DuMate-DeepResearch establishes new state-of-the-art results: the best overall score (58.03%) on DeepResearch Bench, and the best overall score (61.95%) on DeepResearch Bench II while ranking first in information recall and analysis.

Abstract

Explore similar work

DuMate-DeepResearch: An Auditable Multi-Agent System with Recursive Search and Rubric-Grounded Reasoning

S1-DeepResearch: Beyond Search, Toward Real-World Long-Horizon Research Agents

DeepResearch Agent System