AI-Scientist-v2 · Issues· 78 open

Two v2 bugs (F11 VLM whitelist rejection, F13 UnboundLocalError on pdf_path) reproduce in 24/24 biomedical-prompt runs at SHA 96bd51617

Error in using --load-code

`find_pdf_path_for_review` raises UnboundLocalError when no reflection PDF exists, crashing the run at the final review step

Stage summaries are only persisted after all 4 stages finish — interrupted runs leave complete journals but no summaries, blocking the writeup stage

Plot aggregator asks the LLM to guess experiment_data.npy structure → empty figure sets that 5 reflection rounds cannot fix; attach a real data-structure manifest to the prompt

Writeup: retries delete previously produced PDFs, the returned success flag is unreliable, and the final reflection can leave a corrupted template.tex that no longer matches the emitted PDF

End-of-run cleanup SIGTERMs unrelated processes machine-wide (substring match on "python"/"torch"/"mp" cmdlines)

Tree search: a single failed worker node aborts the entire run (`raise` in result-collection loop of `ParallelAgent.step`)

RFC: optional add-ons for reproducibility, token cost, and cross-run memory (happy to PR in small pieces)

Order of experimentation stages

Feature request: official Claude Code / Codex compatibility

“Kepemimpinan Perempuan dalam TPQ Nurul Karomah, Umbulsari Jember: Agama sebagai Infrastruktur Budaya dan Arena Negosiasi Gender”

求问如何使用本地部署的大模型

Error HTTPError: 403 Client Error: Forbidden for url: https://api.semanticscholar.org/graph/v1/paper/search?query=

Agentic tree search for scientific discovery!