Unverified50% confidenceFactExact time
Reference adherence refers to how faithfully a model follows the input reference material during generation, and it is one of the key metrics for measuring the controllability of video generation models.
1
Sources
50%
Confidence
Long-term
Relevance
11/5/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedPrompt adherence is one of the core metrics for measuring the quality of text-to-image models, referring to the degree to which the model's generated images accurately reflect all elements in the user's text description.67% similarUnverifiedReference-anchored evaluation means first fixing a human-annotated gold answer, then having the automatic scorer align to that answer rather than letting the model exercise free discretion.63% similarUnverifiedThe above FPV-CRASH benchmark results use a non-source-separated protocol, in which the same source video may appear in both the training and test splits, so the numbers should be regarded as an upper-bound reference for the reference implementation rather than proof of cross-domain generalization ability.63% similarUnverifiedReference calibration is a one-time preprocessing step in the KAPPO workflow: before formal training, the reference model is used to estimate each prompt's baseline difficulty or baseline reward level over the task distribution, providing a normalized baseline for subsequent prompt-adaptive regularization.62% similarUnverifiedReference-based evaluation paradigms include constructing domain-expert-annotated "claim-evidence" pairs, with commonly used means including citation recall and precision, claim-level factual F1, and faithfulness metrics.62% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1140279API
curl https://kongchang.com/api/v1/knowledge/claims/1140279MCP
get_claim(id=1140279)