LLaVA-1.5-7B
An open-source vision-language model based on the LLaMA architecture that uses the SentencePiece tokenizer, widely used in multimodal hallucination evaluation research.
Timeline (last 90 days)
The adversarial split F1 measured on LLaVA-1.5-7B is 82.21, 1.21 to 8.67 points higher than the three adversarial baselines compared against (81.00, 79.30, 73.54).
Across 9,000 greedy-decoding questions covering all three splits, token ID 3582 (corresponding to the bare word 'yes') and 1217 (corresponding to the bare word 'no') were never once generated by LLaVA-1.5-7B; what the model actually outputs is ' Yes' (token 3869) and ' No' (token 1939).
After switching to an eight-token lookup table covering the space-prefixed forms, LLaVA-1.5-7B achieves F1 of 0.8221, 0.8498, and 0.8713 on the adversarial, popular, and random splits respectively; fixing the reading alone moves the adversarial split by 6.13 points.
Of nine inference-time correction methods tried on a fixed baseline, none surpassed the baseline.
All Facts (4)
The adversarial split F1 measured on LLaVA-1.5-7B is 82.21, 1.21 to 8.67 points higher than the three adversarial baselines compared against (81.00, 79.30, 73.54).
50%UnverifiedAcross 9,000 greedy-decoding questions covering all three splits, token ID 3582 (corresponding to the bare word 'yes') and 1217 (corresponding to the bare word 'no') were never once generated by LLaVA-1.5-7B; what the model actually outputs is ' Yes' (token 3869) and ' No' (token 1939).
50%UnverifiedAfter switching to an eight-token lookup table covering the space-prefixed forms, LLaVA-1.5-7B achieves F1 of 0.8221, 0.8498, and 0.8713 on the adversarial, popular, and random splits respectively; fixing the reading alone moves the adversarial split by 6.13 points.
50%UnverifiedOf nine inference-time correction methods tried on a fixed baseline, none surpassed the baseline.
50%