specific method · filed under training objective
KL alignment loss
Uses a KL-divergence objective to train or supervise the Index Branch by matching its selection distribution.
Also called KL-divergence loss, KL divergence loss for indexer, KL loss.
- sources
- 3
- models
- 2
- labs adopt it
- 2
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
Documented in
Further reading
Picked by hand, not extracted: where to read more, not evidence for anything on this page.
- DeepSeek-V3.2-Exp Technical Report paper github.comthe KL-divergence objective that trains the lightning indexer (eq. 3)
Evidence
4 spans quoted from the sources, strongest treatment first.
train the Index Branch with a KL alignment loss
KL loss directly supervises the Index Branch by matching its selection distribution
Based on p_t,:, we set a KL-divergence loss as the training objective of the indexer
Based on , we set a KL-divergence loss as the training objective of the indexer
Filed alongside
Other methods under training objective :: auxiliary loss.