跳到正文
arXiv cs.AI· Hansang Lee·· 23 小时前AI 评分31

偏好优化让 LLaVA-1.5-7B 少说也少认:诊断并恢复目标覆盖度

Say Less, Hallucinate Less? Diagnosing and Restoring Object Coverage in Preference-Optimized Vision-Language Models

AI 导读

研究发现 LLaVA-1.5-7B 的四个公开偏好优化(PO)检查点在降低目标幻觉的同时也降低了目标覆盖度,命名的目标类别更少,幻觉率下降未带来 CHAIR 集合级目标 F1 的显著提升。

正文 · 原文

View PDF HTML (experimental)

Abstract:Preference optimization (PO) is the standard way to reduce object hallucination in large vision-language models. We find that four public PO checkpoints of LLaVA-1.5-7B also lower object coverage. They name fewer object categories, and their lower hallucination rates bring no significant gain in set-level object F1 on CHAIR. We therefore ask how much of the PO gain the base model can reproduce by stopping earlier. A positive end-of-sequence (EOS) logit bias lets the base model reach the recall of each PO model, and this early stopping alone reproduces 58 to 86 percent of the CHAIRi gain of three checkpoints on two disjoint COCO sets. Writing longer does not bring the lost categories back. Under greedy decoding, a negative EOS bias recovers only 2 to 9 percent of the lost CHAIR recall of these checkpoints, and 94 to 96 percent of the added object mentions repeat categories already named. On CHAIR, the base caption of the same image still names about half of the missed object categories. We therefore propose base-guided object-category restoration, which appends high-confidence base-only categories to the PO caption without training. With thresholds fixed on a development set, it brings three checkpoints within one point of the base recall on a held-out COCO set with a lower distinct-object hallucination rate than the base model. Against the recall-matched base stopping curve, its gain holds for RLAIF-V near the base recall and for RLAIF-V and CHAIR-DPO at an intermediate recall. We will release the code of the recall-controlled evaluation.
Comments: 18 pages, 6 figures, 19 tables. Main text and appendices
Subjects: Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
ACM classes: I.2.10; I.2.7
Cite as: arXiv:2611.23179 [cs.CV]
  (or arXiv:2611.23179v1 [cs.CV] for this version)
  https://doi.org/10.48550/arXiv.2611.23179

arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Hansang Lee [view email]
[v1] Tue, 17 Nov 2026 18:24:58 UTC (355 KB)

来源:arXiv cs.AI · arxiv.org