Proposes dynamic token re-weighting during target-domain fine-tuning to mitigate exacerbated attention sink in source-free CDFSL, achieving SOTA on four benchmarks.
arXiv preprint arXiv:2501.12206 , year=
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
fields
cs.CV 2years
2026 2representative citing papers
A training-free method that steers specific attention heads toward caption-query states reduces object hallucination in LVLMs by ~6% on average across five benchmarks.
citing papers explorer
-
Addressing Exacerbated Attention Sink for Source-Free Cross-Domain Few-Shot Learning
Proposes dynamic token re-weighting during target-domain fine-tuning to mitigate exacerbated attention sink in source-free CDFSL, achieving SOTA on four benchmarks.
-
CAST: Mitigating Object Hallucination in Large Vision-Language Models via Caption-Guided Visual Attention Steering
A training-free method that steers specific attention heads toward caption-query states reduces object hallucination in LVLMs by ~6% on average across five benchmarks.