Research map / Vision-language and multimodal / Multimodal LLMs
TruthPrInt: Mitigating Large Vision-Language Models Object Hallucination Via Latent Truthful-Guided Pre-Intervention
Jinhao Duan, Fei Kong, Hao Cheng, James Diffenderfer, Bhavya Kailkhura, Lichao Sun, Xiaofeng Zhu, Xiaoshuang Shi, Kaidi Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| apply TruthPrInt truthful-guided pre-intervention | MiniGPT-4 greedy decoding · COCO CHAIR | CHAIR_S: 29.53 -> 16.87 | Improved |
| apply TruthPrInt | Llava-1.5 greedy decoding · COCO CHAIR | CHAIR_S: 19.60 -> 10.33 | Improved |
| apply TruthPrInt | mPLUG-Owl2 greedy decoding · COCO CHAIR | CHAIR_S: 23.60 -> 11.13 | Improved |
| apply TruthPrInt | MiniGPT-4 · COCO CHAIR | BLEU: 15.58 -> 17.21 | Improved |
| apply TruthPrInt transferred from MiniGPT-4 detector | InternVL-2.5-8B · COCO CHAIR | CHAIR_S: 13.2 -> 3.2 | Improved |
| apply TruthPrInt transferred from MiniGPT-4 detector | Qwen2-VL-7B-Instruct · COCO CHAIR | CHAIR_S: 12.0 -> 6.2 | Improved |
| apply TruthPrInt | Llava-1.5 · Offline POPE | Precision: 92.80 -> 94.28 | Improved |
| apply TruthPrInt | MiniGPT-4 · original POPE (avg) | Precision: 60.74 -> 61.03 | No clear effect |
| increase max traceback N_B from 1 to 3 (ablation) | TruthPrInt · COCO CHAIR (MiniGPT-4) | CHAIR_S: 16.20 -> 15.20 | Improved |
| increase max traceback N_B from 3 to 5 (ablation) | TruthPrInt · COCO CHAIR (MiniGPT-4) | CHAIR_S: 15.20 -> 15.40 | No clear effect |
| use middle layer 14 instead of layer 12 hidden states (ablation) | TruthPrInt · COCO CHAIR (MiniGPT-4) | CHAIR_S: 24.20 -> 10.40 | Improved |
| use layer 20 instead of layer 14 hidden states (ablation) | TruthPrInt · COCO CHAIR (MiniGPT-4) | CHAIR_S: 10.40 -> 11.80 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- Nullu — Nullu: Mitigating Object Hallucinations in Large Vision-Language Models via HalluSpace Projection (CVPR 2025) · better on 6 of 6 shared results
- OPERA — OPERA: Alleviating Hallucination in Multi-Modal Large Language Models via Over-Trust Penalty and Retrospection-Allocation (CVPR 2024) · better on 4 of 4 shared results
- HALC — HALC: Object Hallucination Reduction via Adaptive Focal-Contrast Decoding (ICML 2024) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Reducing Hallucinations in Large Vision-Language Models via Latent Space Steering (ICLR 2025 · similarity 0.93)
- Analyzing and Mitigating Object Hallucination: A Training Bias Perspective (AAAI 2026 · similarity 0.93)
- Analyzing and Mitigating Object Hallucination in Large Vision-Language Models (ICLR 2024 · similarity 0.93)
- Mitigating Object Hallucination in Large Vision-Language Models via False Discovery Controlled Visual Data Splitting (NeurIPS 2026 · similarity 0.93)
- Generate, but Verify: Reducing Hallucination in Vision-Language Models with Retrospective Resampling (NeurIPS 2025 · similarity 0.93)
- Causally-Grounded Dual-Path Attention Intervention for Object Hallucination Mitigation in LVLMs (AAAI 2026 · similarity 0.93)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
