Chosen and rejected, per pair
Every rollout pair comes back with the rater’s pick, ready to use as a reward signal or as a DPO pair.
How it works
One loop. Live human feedback.
Rollouts go out through the Datapoint API, real people judge them, and rewards stream back while the policy is still training. Low-trust answers are dropped before they ever reach the update.
Reward model
v8updated
reward0.84
Rollouts →← Preferences
API
API840streamed · 10,000+/min
Real people, worldwide1B+ raters · 200+ countries
- 01Your model generates rollouts
- 02Real people judge them
- 03Rewards stream back while it trains
- 04The reward model updates, and keeps going
10,000+human preferences per minute
1B+raters on tap
200+countries represented
Why online
Batch labelling trains on last month’s model
By the time a labelling round comes back, the policy that produced those samples is gone. Online RL needs feedback on what the model is doing now, at the speed it learns.
Batch labellingOnline with Datapoint
Feedback latency
BatchDays to weeks per labelling round
Seconds per judgment, streamed as they land
Which samples get judged
BatchOutputs of a checkpoint the policy has already moved past
The rollouts your policy produced this step, so feedback stays on-policy
Update granularity
BatchOne reward-model refresh per batch
Human reward on every policy update
Reward hacking
BatchFound after the run, in the eval
Caught while it trains, by people a proxy reward cannot stand in for
Who judges
BatchA fixed annotator workforce
Everyday people, targeted by locale, language, age or profession
What comes back
More than a thumbs up
Each preference arrives as structured data your training loop can use directly, with the quality work already done.
Rationales, not just clicks
Raters can write why they chose. Use it to debug the policy or to train a reward model that reasons.
Low-trust answers dropped in flight
Attention checks, response-time floors, agreement scoring and device verification run before a preference reaches your update.
Agreement on every datapoint
Per-datapoint agreement stats tell you which preferences are clear-cut and which are a coin flip, so you can weight them.
Images, audio & video
Compare, rate or rank image, audio and video rollouts on the same API, and mix them within one run.
Elastic, pay per response
Scale the stream up for a sweep and down between runs. No panel to manage, no workforce to contract.