For Collaborators
Living 中文
If you’re a researcher from another field considering working with us, here is how we collaborate.
What we can offer
- Reformulating your problem into a form that can be analyzed and researched
- Based on that analyzable form, methods that are as theoretically grounded as possible and that stand out on small-scale experiments
- Our perspective on existing methods as well (for example: their theoretical strengths and weaknesses, what considerations are missing)
- Open-source, reproducible implementations — not models that exist only in the paper
- An honest statement of capability boundaries — we will tell you plainly what can’t be done yet
What we need from you
- A problem whose solution you actually care about — not a problem that merely happens to have data
- A definition of “what counts as a good result.” We must settle this before starting, or six months in we’ll discover we’ve been chasing different things
- A point of contact willing to join discussions regularly. A collaboration cannot run on one meeting per quarter
Timeline realities
- Methods research moves slower than applied research, with higher uncertainty. We won’t make promises we can’t keep.
- The first meaningful real-world result usually takes at least 3 months, and the first two months mostly go into defining the problem clearly and trying small-scale tests — that stretch doesn’t look like a big output, but it determines everything after.
Authorship
- Settle authorship and data usage before the collaboration starts, in writing. This protects both sides. For how author ordering works inside the lab, see Authorship and Research Contribution.