# Decision-Centric Research: A Calibrated Decision Model and Benchmark for Scientific Agents

DevFeed: [Decision-Centric Research: A Calibrated Decision Model and Benchmark for Scientific Agents](<https://devfeed.tech/articles/decision-centric-research-a-calibrated-decision-model-and-benchmark-for-scientific-agents-115088.md>)

Original publisher: [Read original article](<https://arxiv.org/abs/2611.00774>)

Author: Shaoning Zeng, Wen Huang, Xinzhong Zhu, Yi Duan, Jianhang Zhou, Jianping Gou

Published: 2026-11-03T05:00:00Z

Content type: article

Language: en

Sources: [arXiv: Artificial Intelligence](<https://devfeed.tech/sources/arxiv-artificial-intelligence.md>)

Topics: [Calibrated uncertainty](<https://devfeed.tech/topics/calibrated-uncertainty.md>), [Benchmark](<https://devfeed.tech/topics/benchmark.md>), [generalization in machine learning](<https://devfeed.tech/topics/generalization-in-machine-learning.md>), [AI for science](<https://devfeed.tech/topics/ai-for-science.md>)

Tags: [agents](<https://devfeed.tech/tags/agents.md>), [benchmark](<https://devfeed.tech/tags/benchmark.md>), [calibrated-decision-model](<https://devfeed.tech/tags/calibrated-decision-model.md>), [cs-ai](<https://devfeed.tech/tags/cs-ai.md>), [model](<https://devfeed.tech/tags/model.md>), [research](<https://devfeed.tech/tags/research.md>)

## AI overview

Decision-Centric Research (DCR) evaluates scientific agents through typed choices over an explicit Research State, making individual decisions the unit of supervision and assessment. Its benchmark, RD-Bench, contains 2,000 rule-generated cases covering ten of DCR's 24 decision types. On a held-out research-domain split, the 0.6B state-aware DCRM scored 89.4% Top-1 accuracy, versus 27.8% for a question-only version trained with the same recipe and 36.4% for DeepSeek-V4.1-Flash. After held-out calibration, its prediction sets averaged 1.515 candidates with 97.1% empirical label coverage, 57.4% smaller than the question-only model's sets at similar coverage. These results suggest the explicit state improves decision prediction and provides a compact uncertainty interface; the evidence describes benchmark results, not broader deployment performance.

## Source excerpt

arXiv:2611.00774v1 Announce Type: new Abstract: Scientific agents make consequential choices about evidence, next actions and research strategy, but choices embedded in generated text are difficult to evaluate independently. Decision-Centric Research (DCR) is introduced to make individual choices the common unit of supervision, prediction and evaluation through typed predictions over an explicit Research State. DCR defines 24 decision types, ten of which are instantiated in the Research Decision Benchmark (RD-Bench) of 2,000 rule-generated cases. The Decision-Centric Research Model (DCRM) shares a candidate-token readout across types to produce answers and option probabilities in one forward pass. Under a held-out research-domain split, a 0.6B DCRM (state-aware) achieves 89.4% Top-1 accuracy, compared with 27.8% for DCRM (question-only) trained with the same recipe and 36.4% for DeepSeek-V4.1-Flash. All 30 type-seed comparisons against DCRM (question-only) are significant after Holm correction. After held-out calibration, the resulting prediction sets contain a mean of 1.515 candidates at 97.1% empirical label coverage, reducing mean set size by 57.4% relative to DCRM (question-only) at similar coverage. These results establish a compact state-aware decision layer with accurate typed prediction and an inspectable uncertainty interface.