Sci-MMR: Benchmarking Multi-Step Evidence-Grounded Scientific Reasoning in Multimodal Agents

arXiv 2026 33.5 method

TLDR

Introduces Sci-MMR, a multimodal benchmark for evidence-grounded scientific reasoning, revealing a gap between answer accuracy and evidence recovery in frontier models.

Reasoning

The paper addresses an important gap by evaluating evidence grounding rather than final answers, with a structured benchmark and controlled interventions. However, the abstract is truncated and the scope (235 tasks) may limit generalizability.

Read-first score

Read-first score 33.5, weighted from topical fit, citation, graph, method, reproducibility, and recency signals. Original total remains 15.

Recency 8%

100

Uses a gentle age decay so recent papers surface without erasing older foundations. 2026

Methodology quality 25%

Screens visible abstract and analysis fields for experiment, dataset, baseline, metric, and limitation evidence. markers=benchmark,evaluation,experiment

Reproducibility 25%

Screens links and visible text for paper, code, dataset, artifact, and repository signals. pdf=True; code=False; dataset=False; markers=none

Topical relevance 42%

12.5

Uses existing LLM keyword relevance scores normalized to 0-100. AI scientist,automated scientific discovery,autonomous research agent,automated research,literature review agent,survey generation,automated experimentation,experiment design agent,AI for scientific research,paper writing agent,research automation,scientific discovery agent

Field roles

Frontier

Rank sensitivity

Stability: volatile; rank range: 32.

Keyword Scores

AI for scientific research

autonomous research agent

automated research

research automation

AI scientist

automated scientific discovery

literature review agent

scientific discovery agent

survey generation

automated experimentation

experiment design agent

paper writing agent

Resources

Paper