Share your thoughts, 1 month free Claude Pro on usSee more

Hallucination and Visual Reasoning Evaluation on HallusionBench

81.8Accuracy (aACC)

PStar

Updated 14d ago

Evaluation Results

Method
PStar 2026.05	81.8	33.6	42.8	-	-	-	-	52.7
PStar 2026.05	81.4	36.1	41.1	-	-	-	-	52.9
PStar 2026.05	75.7	23	31.7	-	-	-	-	43.4
Debate with images 2025.11	75.3	58.26	53.79	-	-	-	-	-
Majority vote 2025.11	73.6	56.96	49.46	-	-	-	-	-
Debate about images 2025.11	73.1	56.09	49.82	-	-	-	-	-
Debate with images 2025.11	70.9	50.43	42.24	-	-	-	-	-
Qwen2.5-VL-7B 2026.05	69.9	41.3	44.4	-	-	-	-	51.9
Debate about images 2025.11	69.37	48.26	41.52	-	-	-	-	-
Debate with images 2025.11	69.2	47.39	42.24	-	-	-	-	-
Qwen2-VL-7B 2026.05	68.5	39	43.7	-	-	-	-	50.4
Debate about images 2025.11	68.35	46.96	40.43	-	-	-	-	-
Majority vote 2025.11	68.19	44.35	41.16	-	-	-	-	-
Debate with images 2025.11	67.68	44.78	42.24	-	-	-	-	-
CoT prompt 2025.11	67.17	51.3	40.07	-	-	-	-	-
Majority vote 2025.11	67.17	47.83	39.71	-	-	-	-	-
Debate about images 2025.11	67.17	43.91	39.35	-	-	-	-	-
Direct prompt 2025.11	67.01	39.57	37.91	-	-	-	-	-
Direct prompt 2025.11	66.5	47.39	38.99	-	-	-	-	-
CoT prompt 2025.11	66.5	47.39	41.52	-	-	-	-	-
Direct prompt 2025.11	66.16	49.57	40.07	-	-	-	-	-
CoT prompt 2025.11	65.48	47.83	38.27	-	-	-	-	-
InternVL2-8B 2026.05	63.9	35	36	-	-	-	-	45
GPT-4V (0409) 2026.05	62.1	36.7	32.7	-	-	-	-	43.9
CoT prompt 2025.11	61.76	43.91	35.38	-	-	-	-	-
Qwen2-VL-2B 2026.05	61.5	32.4	33	-	-	-	-	42.4
Direct prompt 2025.11	60.07	41.74	34.3	-	-	-	-	-
InternVL2-2B 2026.05	58.7	26	29.2	-	-	-	-	38
LLaMA-3.2-11B-Vision 2026.05	58	31.8	31.2	-	-	-	-	40.3
Majority vote 2025.11	57.36	21.3	35.38	-	-	-	-	-
LLaVA-v1.5-7B 2026.05	48.8	20.5	13.6	-	-	-	-	27.6
SELF-FILTER (Scores) 2024.02	48.15	11.54	11.11	-	-	-	-	-
SELF-FILTER (CLIP) 2024.02	45.85	14.45	10.33	-	-	-	-	-
SELF-FILTER (w/o Diversity) 2024.02	45.53	10.98	10.99	-	-	-	-	-
Full data 2024.02	45.32	14.74	10.33	-	-	-	-	-
Prototypicality 2024.02	45.32	13.29	10.77	-	-	-	-	-
Random 2024.02	43.85	14.16	9.01	-	-	-	-	-
Alpagasus 2024.02	42.69	10.12	6.81	-	-	-	-	-
GraNd 2024.02	42.27	9.83	8.57	-	-	-	-	-
EL2N 2024.02	41.22	9.25	7.25	-	-	-	-	-
InstructBLIP 2024.05	-	-	-	31.2	-	-	-	-
MiniGPT-4-v1 2024.05	-	-	-	31.9	-	-	-	-
MiniGPT-4-v2 2024.05	-	-	-	30	-	-	-	-
Idefics-instruct 2024.05	-	-	-	27.3	-	-	-	-
OpenFlamingo v2 2024.05	-	-	-	29.4	-	-	-	-
Qwen-VL 2024.05	-	-	-	29.9	-	-	-	-
Qwen-VL-Chat 2024.05	-	-	-	36.8	-	-	-	-
LLaVA-v1.5 2024.05	-	-	-	27.6	-	-	-	-
VLORA 2024.05	-	-	-	26.4	-	-	-	-
Qwen-VL-Max-0809 2024.10	-	-	-	59.2	-	-	-	-
GPT-4o-20240806 2024.10	-	-	-	54.2	-	-	-	-
InternVL2-Llama3-76B 2024.10	-	-	-	55.4	-	-	-	-
Claude3.5-Sonnet 2024.10	-	-	-	49.9	-	-	-	-
Gemini-1.5-Pro 2024.10	-	-	-	45.6	-	-	-	-
GPT-4v (1106, detail-high) 2024.10	-	-	-	39.3	-	-	-	-
Gemini-1.0-Pro 2024.10	-	-	-	45.7	-	-	-	-
Claude3-Sonnet 2024.10	-	-	-	41.3	-	-	-	-
Qwen-VL-Plus 2024.10	-	-	-	40.6	-	-	-	-
Qwen2-VL-2B 2024.10	-	-	-	42.4	-	-	-	-
H2OVL-Mississippi-2B 2024.10	-	-	-	36.4	-	-	-	-
InternVL2-2B 2024.10	-	-	-	38	-	-	-	-
Phi-3-Vision 2024.10	-	-	-	39	-	-	-	-
MiniMonkey 2024.10	-	-	-	30.9	-	-	-	-
MiniCPM-V-2 2024.10	-	-	-	36.1	-	-	-	-
InternVL2-1B 2024.10	-	-	-	34.3	-	-	-	-
PaliGemma-3B-mix-448 2024.10	-	-	-	32.2	-	-	-	-
H2OVL-Mississippi-0.8B 2024.10	-	-	-	29.6	-	-	-	-
DeepSeek-VL-1.3B 2024.10	-	-	-	27.6	-	-	-	-
MiniGPT4 2025.03	-	-	-	31.9	-	-	-	-
Kosmos-2 2025.03	-	-	-	19.8	-	-	-	-
Idefics 2025.03	-	-	-	27.3	-	-	-	-
LLaVA 2025.03	-	-	-	21.6	-	-	-	-
Qwen-VL 2025.03	-	-	-	29.9	-	-	-	-
Emu2-Chat 2025.03	-	-	-	29.5	-	-	-	-
Kosmos-G 2025.03	-	-	-	20.4	-	-	-	-
Chameleon-7b 2025.03	-	-	-	17.1	-	-	-	-
WeGen 2025.03	-	-	-	30.4	-	-	-	-
Qwen2.5VL-7B 2025.12	-	-	-	-	42.4	-	-	-
Qwen2.5VL-7B 2025.12	-	-	-	-	38.6	-	-	-
Video-R1 2025.12	-	-	-	-	40.1	-	-	-
Video-R1 2025.12	-	-	-	-	25.8	-	-	-
Argos 2025.12	-	-	-	-	46.6	-	-	-
R1-OneVision-7B 2026.03	-	-	-	-	65.6	-	-	-
OpenVLThinker-7B 2026.03	-	-	-	-	70.8	-	-	-
MM-Eureka-Qwen-7B 2026.03	-	-	-	-	68.3	-	-	-
ADORA-7B 2026.03	-	-	-	-	53.1	-	-	-
ThinkLite-7B-VL 2026.03	-	-	-	-	71	-	-	-
VLAA-Thinker-Qwen2.5-7B 2026.03	-	-	-	-	68.6	-	-	-
Qwen2.5-VL-7B-Instruct 2026.03	-	-	-	-	71.2	-	-	-
Qwen2.5-VL-7B-Instruct + Vanilla GRPO 2026.03	-	-	-	-	69.8	-	-	-
Qwen2.5-VL-7B-Instruct + NoisyRollout 2026.03	-	-	-	-	70.8	-	-	-
Qwen2.5-VL-7B-Instruct + R-TAP 2026.03	-	-	-	-	77.9	-	-	-
Phi-4-reasoning-vision-15B 2026.03	-	-	-	-	64.4	-	-	-
Phi-4-reasoning-vision-15B 2026.03	-	-	-	-	63.1	-	-	-
Phi-4-mm-instruct 2026.03	-	-	-	-	56	-	-	-
Kimi-VL-A3B-Instruct 2026.03	-	-	-	-	65.2	-	-	-
gemma-3-12b-it 2026.03	-	-	-	-	65.3	-	-	-
Qwen3-VL-8B-Instruct 2026.03	-	-	-	-	73.5	-	-	-
Qwen3-VL-8B-Instruct 2026.03	-	-	-	-	74.1	-	-	-
Qwen3-VL-32B-Instruct 2026.03	-	-	-	-	74.4	-	-	-

Showing 100 of 135 rows