# Multimodal Financial Question Answering on FinVQA multimodal Source: [Multimodal Financial Question Answering on FinVQA multimodal]() Collection index: [llms.txt]() Data updated: 2026-05-14T06:14:54.274Z Evaluation of model performance on the FinVQA multimodal dataset for financial question answering, reporting language-specific accuracy metrics. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Multimodal Financial Question Answering | | Dataset (including any stated split) | FinVQA multimodal | | Metric | Accuracy (en) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 27 of 27 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [Accuracy (en)](): Higher is better; unit: PERCENT. - [Accuracy (hi)](): Higher is better; unit: PERCENT. - [Accuracy (bn)](): Higher is better; unit: PERCENT. - [Accuracy (mr)](): Higher is better; unit: PERCENT. - [Accuracy (gu)](): Higher is better; unit: PERCENT. - [Accuracy (ta)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy (en) | Result date | Report paper | | --- | --- | --- | --- | | Qwen3-VL-32B-Instruct | 69.4 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 69.3 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-32B-Instruct | 68.3 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-32B-Instruct | 67.2 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 66.7 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-8B-Instruct | 64.6 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-32B-Instruct | 64.1 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-32B-Instruct | 62.4 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-4B-Instruct | 61.9 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-7B-Instruct | 61.4 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 61.1 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 59.3 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 59.3 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-32B-Instruct | 58.7 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-8B-Instruct | 58.2 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-8B-Instruct | 58.2 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 57.1 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-3B-Instruct | 55.6 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-4B-Instruct | 55.6 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-3B-Instruct | 54.5 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen3-VL-4B-Instruct | 54.5 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-7B-Instruct | 54.5 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 53.4 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-7B-Instruct | 53.4 | Not recorded | 6a055748dbc836b87d828f1d | | Qwen2.5-VL-3B-Instruct | 52.4 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 51.9 | Not recorded | 6a055748dbc836b87d828f1d | | Gemma-3-4b-it | 37.6 | Not recorded | 6a055748dbc836b87d828f1d | ### Qwen3-VL-32B-Instruct Accuracy (en): 69.4. Result ID: 6a05681147f34ce924ebaae4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"27-32B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 69.3. Result ID: 6a05681147f34ce924ebaae5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Instruct Accuracy (en): 68.3. Result ID: 6a05681147f34ce924ebaae2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B-Instruct Accuracy (en): 67.2. Result ID: 6a05681147f34ce924ebaadf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"27-32B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 66.7. Result ID: 6a05681147f34ce924ebaadc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"7-12B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy (en): 64.6. Result ID: 6a05681147f34ce924ebaad9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"7-12B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B-Instruct Accuracy (en): 64.1. Result ID: 6a05681147f34ce924ebaae1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen3-VL-32B-Instruct Accuracy (en): 62.4. Result ID: 6a05681147f34ce924ebaae3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Instruct Accuracy (en): 61.9. Result ID: 6a05681147f34ce924ebaad0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B-Instruct Accuracy (en): 61.4. Result ID: 6a05681147f34ce924ebaad6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"7-12B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 61.1. Result ID: 6a05681147f34ce924ebaae7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"27-32B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 59.3. Result ID: 6a05681147f34ce924ebaadd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"7-12B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 59.3. Result ID: 6a05681147f34ce924ebaae6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-32B-Instruct Accuracy (en): 58.7. Result ID: 6a05681147f34ce924ebaae0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"27-32B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy (en): 58.2. Result ID: 6a05681147f34ce924ebaada. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"7-12B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy (en): 58.2. Result ID: 6a05681147f34ce924ebaadb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"7-12B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 57.1. Result ID: 6a05681147f34ce924ebaade. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"7-12B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B-Instruct Accuracy (en): 55.6. Result ID: 6a05681147f34ce924ebaacf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Instruct Accuracy (en): 55.6. Result ID: 6a05681147f34ce924ebaad1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B-Instruct Accuracy (en): 54.5. Result ID: 6a05681147f34ce924ebaace. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Instruct Accuracy (en): 54.5. Result ID: 6a05681147f34ce924ebaad2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B-Instruct Accuracy (en): 54.5. Result ID: 6a05681147f34ce924ebaad8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"7-12B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 53.4. Result ID: 6a05681147f34ce924ebaad4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"3-4B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B-Instruct Accuracy (en): 53.4. Result ID: 6a05681147f34ce924ebaad7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Constrained decoding","Scale":"7-12B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B-Instruct Accuracy (en): 52.4. Result ID: 6a05681147f34ce924ebaacd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"3-4B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 51.9. Result ID: 6a05681147f34ce924ebaad5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Supervised finetuning","Scale":"3-4B"}. This record does not assert independent reproduction. ### Gemma-3-4b-it Accuracy (en): 37.6. Result ID: 6a05681147f34ce924ebaad3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Inference Mode":"Zero shot inference","Scale":"3-4B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.