Hi team, and congratulations on AI-Researcher (NeurIPS 2025) — it's an impressive system. I deployed it and have been running it locally.
Two things:
License: I noticed the repository doesn't include a LICENSE file, which (under default copyright) makes it unclear whether the community may build and release derivatives. Would you consider adding an explicit license (e.g., MIT or Apache-2.0)? It would help others contribute and extend the work with confidence.
A contribution back: To run AI-Researcher in environments without Docker (e.g., AutoDL GPU instances, which forbid nested Docker, and shared Slurm clusters), I implemented a drop-in LocalEnv that mirrors DockerEnv's interface but executes agent commands in the local shell — no tool code changes needed (duck-typed on run_command / local_workplace). I'd be happy to open a PR if it's useful to you. (I also hit a couple of bare-install issues on main — a missing PLATFORM in constant.py and some dependency pins — which I can include.)
I'm also interested in extending AI-Researcher toward multimodal / vision-language autoresearch (the current benchmark covers diffu_flow / gnn / reasoning / recommendation / vq, but not multimodal). Before doing anything public, I wanted to ask: would you welcome an attributed, citing derivative in that direction? Happy to coordinate however you prefer.
Thanks for open-sourcing this!
Hi team, and congratulations on AI-Researcher (NeurIPS 2025) — it's an impressive system. I deployed it and have been running it locally.
Two things:
License: I noticed the repository doesn't include a LICENSE file, which (under default copyright) makes it unclear whether the community may build and release derivatives. Would you consider adding an explicit license (e.g., MIT or Apache-2.0)? It would help others contribute and extend the work with confidence.
A contribution back: To run AI-Researcher in environments without Docker (e.g., AutoDL GPU instances, which forbid nested Docker, and shared Slurm clusters), I implemented a drop-in LocalEnv that mirrors DockerEnv's interface but executes agent commands in the local shell — no tool code changes needed (duck-typed on run_command / local_workplace). I'd be happy to open a PR if it's useful to you. (I also hit a couple of bare-install issues on main — a missing PLATFORM in constant.py and some dependency pins — which I can include.)
I'm also interested in extending AI-Researcher toward multimodal / vision-language autoresearch (the current benchmark covers diffu_flow / gnn / reasoning / recommendation / vq, but not multimodal). Before doing anything public, I wanted to ask: would you welcome an attributed, citing derivative in that direction? Happy to coordinate however you prefer.
Thanks for open-sourcing this!