Per-instance adaptive projection onto clustered hallucination subspaces reduces LVLM hallucination on CHAIR and POPE benchmarks without fine-tuning.
Foundation model-powered 3d few- shot class incremental learning via training-free adaptor
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
citation-role summary
background 1
citation-polarity summary
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1roles
background 1polarities
unclear 1representative citing papers
citing papers explorer
-
Beyond Global Editing: Per-Instance Disentangled Subspaces for Training-Free Hallucination Mitigation in LVLMs
Per-instance adaptive projection onto clustered hallucination subspaces reduces LVLM hallucination on CHAIR and POPE benchmarks without fine-tuning.