ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Tasks

ResearchBench is the official dataset for the ACL Findings 2026 paper. It provides benchmark data for three inspiration-based subtasks of scientific discovery, including inspiration retrieval. The dataset was released by author ankilok and last updated on May 31, 2026.

Use Cases

• Benchmarking LLM performance on inspiration retrieval tasks based on the described subtask.
• Evaluating scientific discovery capabilities of models based on inspiration-based task decomposition.
• Reproducing results from the ACL Findings 2026 paper based on the provided benchmark data.

Strengths

• Official dataset for a peer-reviewed ACL Findings 2026 paper.
• Includes evaluation code, CLI tools, and reproduction instructions via a linked GitHub repository.
• Last updated on May 31, 2026, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and data size are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

ankilok

Collection Method

Released as part of the ACL Findings 2026 paper 'ResearchBench: Benchmarking LLMs in Scientific Discovery via Inspiration-Based Task Decomposition'.

Freshness

2026-05-31 17:18:41