Unverified50% confidenceFactExact time
The study demonstrates through theoretical analysis that interference from irrelevant visual tokens causes measurable performance degradation in large vision-language models.
1
Sources
50%
Confidence
Long-term
Relevance
9/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedUsing chain-of-thought (CoT) prompting not only failed to improve visual perception performance but actually degraded it.75% similarUnverifiedAccording to Artificial Analysis's evaluation, the vision model ranks slightly lower due to having sacrificed some visual capabilities.75% similarUnverifiedPrior to this study, the academic community rarely conducted systematic quantification of performance degradation caused by irrelevant visual tokens.74% similarUnverifiedVision-language models are relatively mature in evaluating prompt adherence, but still have limitations in dimensions such as motion quality that require temporal understanding.73% similarUnverifiedMultimodal large models can leverage language priors to complete ambiguous characters, greatly diminishing the effectiveness of attacks targeting the visual features of individual characters.71% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/921802API
curl https://kongchang.com/api/v1/knowledge/claims/921802MCP
get_claim(id=921802)