Mitigating Object Hallucination via Concentric Causal Attention

TL;DR

This paper identifies the link between Rotary Position Encoding and object hallucination in LVLMs, and proposes Concentric Causal Attention to improve factual alignment with images.

Contribution

It introduces Concentric Causal Attention, a novel positional alignment method that reduces object hallucination by addressing RoPE's long-term decay effects in LVLMs.

Findings

CCA significantly reduces object hallucination in LVLMs.

CCA outperforms existing mitigation strategies on multiple benchmarks.

The study reveals the impact of RoPE decay on visual-instruction interaction.

Abstract

Recent Large Vision Language Models (LVLMs) present remarkable zero-shot conversational and reasoning capabilities given multimodal queries. Nevertheless, they suffer from object hallucination, a phenomenon where LVLMs are prone to generate textual responses not factually aligned with image inputs. Our pilot study reveals that object hallucination is closely tied with Rotary Position Encoding (RoPE), a widely adopted positional dependency modeling design in existing LVLMs. Due to the long-term decay in RoPE, LVLMs tend to hallucinate more when relevant visual cues are distant from instruction tokens in the multimodal input sequence. Additionally, we observe a similar effect when reversing the sequential order of visual tokens during multimodal alignment. Our tests indicate that long-term decay in RoPE poses challenges to LVLMs while capturing visual-instruction interactions across long distances. We propose Concentric Causal Attention (CCA), a simple yet effective positional alignment strategy that mitigates the impact of RoPE long-term decay in LVLMs by naturally reducing relative distance between visual and instruction tokens. With CCA, visual tokens can better interact with instruction tokens, thereby enhancing model's perception capability and alleviating object hallucination. Without bells and whistles, our positional alignment method surpasses existing hallucination mitigation strategies by large margins on multiple object hallucination benchmarks.