JevGPT: Thinking Fast and Slow in Autonomous Research

Yasheng Sun; Yanning Dai; Yimeng Chen; Nanbo Li; Firas Laakom; Yifan Xu; Piotr Piękos; Mateusz Ostaszewski

cs.CLOct 13, 2026

JevGPT: Thinking Fast and Slow in Autonomous Research

Authors: Yasheng Sun, Yanning Dai, Yimeng Chen, Nanbo Li, Firas Laakom, Yifan Xu, Piotr Piękos, Mateusz Ostaszewski, +4 more

Organizations: King Abdullah University of Science and Technology (KAUST)

Abstract

Autonomous research agents must acquire and reason over evidence within a limited budget. Yet relevant sources can remain uninspected or unused, leaving a gap between retrieval coverage and the evidence that informs research. The challenge is to align fast local judgments with changing evidence needs and feed their findings back into research decisions. We introduce JevGPT, a dual-system architecture for autonomous research that couples fast evidence processing with deliberative research planning. System Two sets evidence objectives and integrates findings, while System One locates support, assesses relevance, and guides further inspection. Independent source judgments run in parallel; selected original passages inform the next research step. Experiments on fixed-corpus question answering and document ranking show that JevGPT achieves leading task performance, with faster completion and lower cost than comparably effective research agents. Evidence analysis shows broader source inspection and greater coverage of supporting documents in the research loop, highlighting efficient evidence processing as a route to more capable autonomous research.

Figures & tables

Figure 2: Quality and API cost. Stars mark JevGPT; upper left is better. Costs include both systems for JevGPT; the NeuCLIRBench panel shows GPT-based methods.

Panel of Figure 3 — Figure 3: Overview of JevGPT. System Two plans research and integrates evidence; System One guides evidence localization, relevance assessment, and discovery. The readers return original passages with local judgments. Dashed loops adapt reading; the dashed discovery branch supplies additional research leads asynchronously.

		Performance	Resources / question			Cost / question ($)
Method	Retriever	Accuracy (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
BCP/ReAct	BM25	$77.11$	$2,169$	$24.87$	$182.1$	$3.0798$	—
Pi-Serini	BM25	$82.05$	266	$14.41$ $14.41$	$112.6$	0.4053	—
Agentica	BM25	$84.46$	$335$ $335$	$33.88$	$111.7$	$0.7919$	—
AtlasNav	BM25 + map	$89.16$	$948$	$20.54$	$144.0$	$1.5257$	—
RISE	BM25	$89.88$	$609$	$26.08$	$111.0$ $111.0$	$1.1050$	—

Table 1: BrowseComp-Plus results on 830 questions. Resources are per-question means; S2/S1 denote the two systems’ API costs. Bold and underlining mark the best and second-best values, including ties.

		Performance		Resources / topic			Cost / topic ($)
Method	Retriever	nDCG@20 $\uparrow$	F 1 @20 (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
BM25	BM25	$0.3608$	$26.13$	—	—	—	—	—
RM3	BM25	$0.3622$	$26.17$	—	—	—	—	—
Patapsco	BM25	$0.3485$	$24.80$	—	—	—	—	—
Qwen3-Reranker	Fusion	$0.5138$	$35.06$ $35.06$	—	—	—	—	—
RankGPT	BM25	$0.4843$	$32.48$	76	1.00	$141.3$ $141.3$	$0.6402$ $0.6402$	—

Table 2: NeuCLIRBench results on 105 topics. Resources are per-topic means. Bold and underlining mark the top two values; dashes denote inapplicable or unreported resources.

Figure 4: Gold-document coverage. Mean per-question recall over 830 BrowseComp-Plus questions, measured separately for each document set under native retrieval settings.

Figure 5: Evidence localization for answer verification. System One locates an explicit statement of Ancrum’s by-name, giving System Two a concrete candidate for its next query. Selected steps are shown.

Configuration	Accuracy (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
	Performance	Resources / question			Cost / question ($)
Pi-Serini	$72.05$	$133$	$9.29$	$76.6$	$0.0895$	N/A
+ Hybrid BM25	$79.52$	118	$8.70$	$72.9$	0.0824	N/A
+ Evidence Localization	$81.20$	$202$	$7.17$	$75.6$	$0.1353$	0.0148
+ Evidence Discovery	86.51	$205$	5.72	70.2	$0.1403$	$0.0448$

Table 3: Cumulative ablation study on BrowseComp-Plus. All configurations use GPT-5.6 Terra on 415 matched questions with a 300-second active research budget; Evidence Discovery uses

K=60

. Bold marks the best value; N/A indicates that System One is unused.

	Performance	Resources / question			Cost / question ($)
System Two	Accuracy (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
GPT-5.2	$83.86$	$255$	$6.54$	$126.9$	$0.1659$	$0.0456$
GPT-5.4	$86.99$	$361$	$10.66$	$165.2$	$0.2869$	$0.0561$
GPT-5.4 mini	$79.76$	$251$	$9.56$	$152.2$	0.0707	$0.0479$
GPT-5.5	92.29	$418$	$8.91$	$92.8$	$0.5955$	$0.0608$
GPT-5.6 Terra	$86.27$	176	$5.62$	82.9	$0.1304$	0.0434

Table 4: System Two model comparisons on BrowseComp-Plus. All models use 415 matched questions and the main elapsed-time budget; the Terra row is a separate run from the last row of Table 3 . Bold marks the best value.

Figure 6: Cost and accuracy across System Two models. Each point represents JevGPT with a different System Two model. API cost includes both systems. The dashed line marks the observed Pareto frontier.

Appendix figures & tables3 assets

Supplementary material from the paper’s appendix.

Appendix

Configuration	nDCG@20 $\uparrow$	F 1 @20 (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
	Performance		Resources / topic			Cost / topic ($)
Pi-Serini	$0.4904$	$31.17$	$162$	$4.76$	$75.4$	$0.1405$	N/A
+ Hybrid BM25	$0.4855$	$30.56$	148	$4.90$	$77.6$	0.1328	N/A
+ Relevance Assessment	0.5256	$35.28$	$169$	2.92	$72.2$	$0.4147$	0.0088
+ Evidence Discovery	$0.5206$	35.63	$151$	$3.12$	61.7	$0.3706$	$0.0094$

Appendix

Table 5: Cumulative ablation study on NeuCLIRBench. GPT-5.6 Terra as System Two; 105 topics. The full configuration is evaluated independently of Table 6 . Time is elapsed time; costs are estimated from recorded token usage at common rates. Bold marks the best value. N/A indicates that System One is unused.

System Two	nDCG@20 $\uparrow$	F 1 @20 (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
	Performance		Resources / topic			Cost / topic ($)
GPT-5.2	$0.5005$	$33.69$	$107$	$1.57$	$68.7$	$0.1446$	$0.0063$
GPT-5.4	$0.4431$	$27.11$	57	$1.60$	$51.2$	$0.1277$	$0.0056$
GPT-5.4 mini	$0.4580$	$29.34$	$104$	1.48	$71.2$	0.0690	0.0054
GPT-5.5	0.5390	36.10	$158$	$2.68$	$79.3$	$0.5756$	$0.0082$
GPT-5.6 Terra	$0.5244$	$35.65$	$160$	$2.91$	44.8	$0.3921$	$0.0094$

Appendix

Table 6: System Two model comparisons on NeuCLIRBench. All models use 105 topics and a 300-second elapsed-time budget. The Terra configuration is evaluated independently of Table 5 . Bold marks the best value.

System One	nDCG@20 $\uparrow$	F 1 @20 (%) $\uparrow$	Tokens (K) $\downarrow$	Searches $\downarrow$	Time (s) $\downarrow$	S2 $\downarrow$	S1 $\downarrow$
	Performance		Resources / topic			Cost / topic ($)
Jev ( $K=0$ )	$0.5491$	36.78	158	2.50	$82.1$	0.5515	0.0074
Jev ( $K=20$ )	$0.5339$	$36.05$	$172$	$2.52$	$77.2$	$0.5894$	$0.0075$
Jev ( $K=60$ )	$0.5358$	$35.82$	$168$	$3.02$	74.8	$0.5607$	$0.0093$
Jev ( $K=120$ )	0.5510	$36.73$	$174$	$4.86$	$97.7$	$0.6072$	$0.0182$
GPT-5.5 ( $K=0$ )	$0.5310$	$35.83$	$169$	$2.61$	$353.3$	$0.5937$	1.7363

Appendix

Table 7: System One models and candidate inspection depth on NeuCLIRBench. Jev and GPT-5.5 are compared as System One, with GPT-5.5 fixed as System Two. All configurations use 105 topics and a 3,600-second research budget. At

K=0

, Evidence Discovery is disabled and requested Relevance Assessment remains enabled. Bold marks the best value within each block.

Explore similar work

Argus: Evidence Assembly for Scalable Deep Research Agents
May 15, 2026Zhen Zhang, Liangcai Su, Zhuo Chen +7Multi-Agent LLM Systems Tool-Augmented Language Model Agents
MiroMind AI
EviGraph: Evidence-Guided Autonomous Research Agents
Aug 5, 2026Zhenjiang Ren, Ruiji Li, Xujing Zhang +3Evidence-Grounded Reasoning Evidence-Grounded Generation
Institute of Automation, Chinese Academy of Sciences · School of Artificial Intelligence, University of Chinese Academy of Sciences · Hong Kong Baptist University +1
Auto Research for Materials: Auditable AI-Scientist Workflows with Held-Out Transfer
Jul 19, 2026Jingjie Ning, Xiaochuan Li, Shanshan Zhong +2Automated Algorithm Discovery Automated Evaluation
School of Computer Science, Carnegie Mellon University · 2DP Technology

May 15, 2026J/K move · Enter open · S save

Argus: Evidence Assembly for Scalable Deep Research Agents

Zhen Zhang, Liangcai Su, Zhuo Chen, Xiang Lin+6

MiroMind AI

Multi-Agent LLM Systems Tool-Augmented Language Model Agents Deep Research Agents Agentic Search Agentic Reasoning

Open paper AskPDF arXiv

Deep research agents have achieved remarkable progress on complex information seeking tasks. Even long ReAct style rollouts explore only a single trajectory, while recent state of the art systems scale inference time compute via parallel search and aggregation. Yet deep research answers are composed of complementary pieces of evidence, which parallel rollouts often duplicate rather than complete, yielding diminishing returns while pushing the aggregation context toward the model's limit. We propose Argus, an agentic system in which a Searcher and a Navigator cooperate to treat deep research as assembling a jigsaw from complementary evidence pieces, rather than brute forcing the whole answer in parallel. The Searcher collects evidence traces for a given sub-query through ReAct-style interaction. The Navigator maintains a shared evidence graph, verifying which pieces are still missing, dispatching Searchers to gather them, and reasoning over the completed graph to produce a source-traced final answer. We train the Navigator with reinforcement learning to verify, dispatch, and synthesize, while independently training the Searcher to remain a standard ReAct agent. The resulting Navigator supports rollouts with a single Searcher or many in parallel without retraining. With both Searcher and Navigator built on a 35B-A3B MoE backbone, Argus gains 5.5 points with a single Searcher and 12.7 points with 8 parallel Searchers, averaged over eight benchmarks. With 64 Searchers it reaches 86.2 on BrowseComp, surpassing every proprietary agent we benchmark, while the Navigator's reasoning context stays under 21.5K tokens.

Abstract

Explore similar work

Argus: Evidence Assembly for Scalable Deep Research Agents

EviGraph: Evidence-Guided Autonomous Research Agents

Auto Research for Materials: Auditable AI-Scientist Workflows with Held-Out Transfer