cs.AIApr 7, 2026

Towards Trustworthy Report Generation: A Deep Research Agent with Progressive Confidence Estimation and Calibration

Authors: Yi Yuan, Xuhong Wang, Shanzhe Lei

Organizations: Shanghai Artificial Intelligence Laboratory · Southeast University

Read PDF
Ask AI
Cite

Copy citation

DOINot available for this paper

Abstract

As agent-based systems continue to evolve, deep research agents are capable of automatically generating research-style reports across diverse domains. While these agents promise to streamline information synthesis and knowledge exploration, existing evaluation frameworks-typically based on subjective dimensions-fail to capture a critical aspect of report quality: trustworthiness. In open-ended research scenarios where ground-truth answers are unavailable, current evaluation methods cannot effectively measure the epistemic confidence of generated content, making calibration difficult and leaving users susceptible to misleading or hallucinated information. To address this limitation, we propose a novel deep research agent that incorporates progressive confidence estimation and calibration within the report generation pipeline. Our system leverages a deliberative search model, featuring deep retrieval and multi-hop reasoning to ground outputs in verifiable evidence while assigning confidence scores to individual claims. Combined with a carefully designed workflow, this approach produces trustworthy reports with enhanced transparency. Experimental results and case studies demonstrate that our method substantially improves interpretability and significantly increases user trust.

Figures & tables

Figure 1: Illustration of overconfidence in question answering. When asked “Who was the first person to walk on the Moon?” , an overconfident LLM might respond with an incorrect but assertive answer such as “Yuri Gagarin” who was the first human in space but never walked on the Moon, which can mislead users. In contrast, a calibrated model may still provide an incorrect answer, but with lower confidence, encouraging users to verify the information.
Figure 2: A three-stage framework for autonomous trustworthy report generation, consisting of a planning module, multiple research workers and synthesis module.
MethodsAccuracy ↑\uparrow
Qwen-VL2.5-72B48.99
GPT-4o33.48
Gemini-2.0-Flash32.39
Gemini-2.0-Flash-Thinking40.91
Gemini-2.5-Pro-Exp52.02
Claude-3-7-Sonnet54.04
Table 1: Accuracy results of various large language models evaluated on the GPQA-Diamond benchmark.
MethodsNECE↓\mathbf{N}_{\mathrm{ECE}}\downarrow
Qwen-VL2.5-72B0.53
Intern-VL3-78B0.50
DeepSeek-R1-Distill-Llama-70B0.42
GPT-4.10.47
GPT-4o0.39
Claude-4-Sonnet0.36
Table 2: Calibration performance measured by the normalized Expected Calibration Error ( NECE\mathbf{N}_{\mathrm{ECE}} ) on the xBench-DeepSearch benchmark across various large language models.
MethodOverallComprehensivenessInsightfulnessInstructionReadability
Deep Research Agents
Gemini-2.5-Pro Deep Research48.9248.4548.3049.2949.77
OpenAI Deep Research46.4546.4643.7349.3947.22
Claude-Researcher45.0045.3442.7947.5844.66
Kimi-Researcher44.6444.9641.9747.1445.59
Doubao-DeepResearch44.3444.8440.5647.9544.69
Table 3: Comparative results of different deep research agents and LLMs with search tools on the Deep Research Bench across four evaluation dimensions.
More in this topic

Deep Research Agents

Deep research agents plan long-horizon searches, synthesize cited evidence, and produce reports for complex open-ended questions.

See all 115 papers

Explore similar work

CardsList
Sep 1, 2026

DualStake: Dual-Path Confidence Calibration in Deep Research Agents

Deep Research agents tackle knowledge-intensive tasks through multi-round retrieval and decision-oriented generation. However, these agents suffer from severe overconfidence, making their expressed confidence unreliable for user trust and downstream abstention. To address this, we augment the Deep Research pipeline with step confidence elicitation after each retrieval, building on the commonly used post-answer verbalized confidence. Interestingly, we find that Evidence Confidence (E-Conf), elicited after the final retrieval step, provides a stronger uncertainty signal than Answer Confidence (A-Conf), elicited after answer generation, and that A-Conf is largely shaped by E-Conf. Based on these findings, we propose DualStake, a dual-path calibration method that applies margin-clipped, confidence-dependent stake rewards to jointly align E-Conf and A-Conf with answer correctness while limiting extreme confidence optimization. Experiments on Qwen2.5-7B, Qwen2.5-7B-Instruct, and Qwen3-4B across 8 QA benchmarks demonstrate that DualStake consistently improves calibration without sacrificing answer accuracy. The code is available at https://github.com/FloXXXt/DualStake.

Yinuo Xu, Yuwei Liang, Jianjie Cheng +4

NLPR & MAIS, Institute of Automation, Chinese Academy of Sciences (CASIA) · School of Artificial Intelligence, University of Chinese Academy of Sciences · Meituan Inc.

Deep Research AgentsLanguage Model CalibrationLLM Uncertainty EstimationUncertainty Calibration
AskPDF
Apr 16, 2026

DR3^{3}-Eval: Towards Realistic and Reproducible Deep Research Evaluation

Deep Research Agents (DRAs) aim to solve complex, long-horizon research tasks involving planning, retrieval, multimodal understanding, and report generation, yet their evaluation remains challenging due to dynamic web environments and ambiguous task definitions. We propose DR3^{3}-Eval, a realistic and reproducible benchmark for evaluating deep research agents on multimodal, multi-file report generation. DR3^{3}-Eval is constructed from authentic user-provided materials and paired with a per-task static research sandbox corpus that simulates open-web complexity while remaining fully verifiable, containing supportive documents, distractors, and noise. Moreover, we introduce a multi-dimensional evaluation framework measuring Information Recall, Factual Accuracy, Citation Coverage, Instruction Following, and Depth Quality, and validate its alignment with human judgments. Experiments with our developed multi-agent system DR3^{3}-Agent based on multiple state-of-the-art language models demonstrate that DR3^{3}-Eval is highly challenging and reveals critical failure modes in retrieval robustness and hallucination control. Our code and data are publicly available.

Qianqian Xie, Qingheng Xiong, He Zhu +16

Nanjing University · M-A-P · Nanjing University of Science and Technology +2

LLM Agent EvaluationLong-Horizon Agent EvaluationAI Agent Benchmarks
AskPDF
Feb 21, 2026

DREAM: Deep Research Evaluation with Agentic Metrics

Deep Research Agents generate analyst-grade reports, yet evaluating them remains challenging due to the absence of a single ground truth and the multidimensional nature of research quality. Recent benchmarks propose distinct methodologies, yet they suffer from the Mirage of Synthesis, where strong surface-level fluency and citation alignment can obscure underlying factual and reasoning defects. We characterize this gap by introducing a taxonomy across four verticals that exposes a critical capability mismatch: static evaluators inherently lack the tool-use capabilities required to assess temporal validity and factual correctness. To address this, we propose DREAM (Deep Research Evaluation with Agentic Metrics), a framework that instantiates the principle of capability parity by making evaluation itself agentic. DREAM structures assessment through an evaluation protocol combining query-agnostic metrics with adaptive metrics generated by a tool-calling agent, enabling temporally aware coverage, grounded verification, and systematic reasoning probes. Controlled evaluations demonstrate DREAM is significantly more sensitive to factual and temporal decay than existing benchmarks, offering a scalable, reference-free evaluation paradigm.

Elad Ben Avraham, Changhao Li, Ron Dorfman +8

AWS Agentic AI · Georgia Institute of Technology

AI Agent EvaluationDeep Research AgentsReference-Free EvaluationTool-Use Evaluation
AskPDF