AI

OmniConfess: Eliciting Token Confessions to Mitigate Omni-Modal Hallucination

Read full article at source ↗

Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 06/10/2026, 05:18:24 UTC.

Briefings

OmniConfess is a training-free method for mitigating Omni-Modal Hallucination in large language models. It fixes a candidate response and re-scores it at token resolution under controlled channel-wise evidence interventions, producing a structured token-by-channel confession that reveals the response's evidential dependence. This approach preserves grounded content and corrects commitments driven by irrelevant or contradictory evidence. Evaluated on OmniHalluBench with 3,540 examples across six datasets, the method significantly reduces hallucinations across heterogeneous modality and task settings.

Key points

• OmniConfess fixes a candidate response and re-scores it at token resolution under controlled channel-wise evidence interventions to produce a structured token-by-channel confession.
• OmniConfess mitigates hallucinations across heterogeneous modality and task settings using a 3,540-example benchmark built from six datasets spanning text, image, audio, and video settings.

Context

OmniLLMs unify text, images, audio, and video, yet hallucinate when generation relies on the wrong evidence. Existing inference-time methods can reduce hallucinations, but rarely reveal which evidence sustains a generated commitment. OmniConfess addresses this by using a structured token-by-channel confession to preserve grounded content and correct commitments driven by irrelevant or contradictory evidence.

Limitations

The evaluation relies on a single publicly available code repository and benchmark; no additional experimental data or validation results are provided in the source evidence.

Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 18/10/2026, 02:21:29 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.

View archived briefing on IPFS
bafybeidpnirtyscbmoyg4jijrfiivma3xpvhulrnvkh4ztdz7gywxjxjai