Open Models Almanac
Methodsdata curationsynthetic data

specific method · filed under data curation

Knowledge distillation for synthetic data generation

Knowledge distillation trains a student model using signal from strong teacher models rather than only from external data.1 In the on-policy variant, the teacher signal is distilled onto the student's own trajectories, aligning training with what the student would actually produce at inference time.2

Trade-off On-policy distillation yields stronger gains than purely off-policy distillation.3

How this definition was written

Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.

  1. This technique uses strong teacher models to guide training on the model's own generated attempts (on-policy rollouts — nvidia/nemotron-3-ultra-550b-a55b
  2. By distilling teacher signal onto the student's own trajectories rather than offline traces, MOPD better aligns the student's behavior with what it would actually produce at inference time — nvidia/nemotron-3-ultra-550b-a55b
  3. yielding stronger gains than purely off-policy distillation — nvidia/nemotron-3-ultra-550b-a55b

Also called knowledge distillation, distillation.

sources
3
models
2
lab adopt it
1
strongest
used

Documented in

Evidence

3 quotes from 3 sources, one card per model, each opening with its strongest.

How sources treat it
used 3

we generate synthetic data by distilling trajectories, solutions, and translations from strong teacher models and agent systems, often grounded in real tasks or documents and aggressively filtered for quality.

useddata curationin Nemotron 3.5 LightningNVIDIA
1 more from this model

During post-training, we generate synthetic data by distilling trajectories, solutions, and translations from strong teacher models and agent systems

useddata curationin Nemotron 3.5 LightningNVIDIA
↑