Core offering · Expert skills data
Tacit knowledge, made model-ready.
Practicing professionals show how they reason. We turn it into tasks, reference answers and rubrics your models can learn from.
Best for: reasoning models, SFT, RLHF and evals from domain experts
Sample record · FinanceIllustrative example
TaskIs an 11.0× exit multiple defensible as a base case, given the deal and peer data provided?
Expert answer · excerptNot as a base case. All five listed peers trade below 11.0×, so I'd underwrite exit at 9.0–9.5× and treat 11.0× as upside.
Rubric5 criteria · 10 pts
View full sample
Material provided
- Entry
- 9.5× EBITDA
- Leverage
- 5.5× net debt / EBITDA
- Model exit
- 11.0× in year 5
- Listed peers
- 5 companies, 8.2–10.1× (median 9.0×)
Full expert answerNot as a base case. All five peers trade below 11.0× (median 9.0×), so the model assumes this company re-rates above every comparable. That could be justified by faster growth, better margins or a scarcity premium, but none of that is in the material. I'd underwrite exit at 9.0–9.5×, in line with the peer median, and show 11.0× as an upside case. At 5.5× leverage, a lower exit multiple also leaves less room for error at refinancing.
Rubric · written by the expert
| Criterion | Points |
|---|---|
| Tests the exit multiple against the peer range provided | 3 |
| States a clear base-case view with reasoning; a different conclusion earns full credit if the evidence supports it | 3 |
| Names what evidence would justify a higher multiple | 2 |
| Connects leverage to the risk of a lower exit multiple | 1 |
| Uses only the figures provided; no unsupported numbers | 1 |
| Total | 10 |
Also from the same expert network
World model data
Game-engine worlds,
captured frame by frame.
RGB + depthCamera posePlayer actionsMulti-view
96.4–98.4s Place the pink object into the first box
Egocentric · robotics
Physical AI data
Raw screen
GUI agents
Computer-use data
4 views
Blender · rubric-graded
3D & spatial data
Unit plan
Architecture