[KongchangAI]
Unverified50% confidenceBenchmarkExact time

In language modeling experiments, using the generation perplexity of Qwen 2.5 1.5B as the oracle reward and BERT, distilled BERT, token entropy, etc. as surrogates, with a feedback set of only 256 values or rankings, the KL between the q_G learned by Martini and q* is significantly lower than that of uniform weighting or Pareto-type methods.

1
Sources
50%
Confidence
Long-term
Relevance
11/2/2026
First Seen

Sources

Related Entities

Related Claims

Cite This Claim

Stable URI
https://kongchang.com/claim/1127430
API
curl https://kongchang.com/api/v1/knowledge/claims/1127430
MCP
get_claim(id=1127430)