# Multi-Image Reasoning on VideoEspresso Source: [Multi-Image Reasoning on VideoEspresso]() Collection index: [llms.txt]() Data updated: 2026-05-28T08:21:47.626Z Evaluation of multi-image reasoning capabilities using the VideoEspresso dataset, reporting performance across 15 specific reasoning dimensions (e.g., Causal, Emotional, Contextual) and an overall average score. 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 | Multi-Image Reasoning | | Dataset (including any stated split) | VideoEspresso | | Metric | Narration Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [Narration Score](): Higher is better; unit: PERCENT. - [Event Score](): Higher is better; unit: PERCENT. - [Ingredient Score](): Higher is better; unit: PERCENT. - [Causal Score](): Higher is better; unit: PERCENT. - [Theme Score](): Higher is better; unit: PERCENT. - [Contextual Score](): Higher is better; unit: PERCENT. - [Influence Score](): Higher is better; unit: PERCENT. - [Role Score](): Higher is better; unit: PERCENT. - [Interaction Score](): Higher is better; unit: PERCENT. - [Behavior Score](): Higher is better; unit: PERCENT. - [Emotional Score](): Higher is better; unit: PERCENT. - [Cooking Score](): Higher is better; unit: PERCENT. - [Traffic Score](): Higher is better; unit: PERCENT. - [Situation Score](): Higher is better; unit: PERCENT. - [Overall Average Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Narration Score | Result date | Report paper | | --- | --- | --- | --- | | ROVER-LSW | 50.6 | Not recorded | 6a17cbf6dbd7150506f8119d | | ROVER-LSW | 49.4 | Not recorded | 6a17cbf6dbd7150506f8119d | | VideoEspresso | 45.2 | Not recorded | 6a17cbf6dbd7150506f8119d | | LongVA-DPO | 35.5 | Not recorded | 6a17cbf6dbd7150506f8119d | | Qwen-VL-Max | 33.9 | Not recorded | 6a17cbf6dbd7150506f8119d | | InternVL2 | 33.9 | Not recorded | 6a17cbf6dbd7150506f8119d | | GPT-4o | 32.3 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-1.5 | 32.3 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-N-Video | 31.2 | Not recorded | 6a17cbf6dbd7150506f8119d | | mPLUG-Owl3 | 30.6 | Not recorded | 6a17cbf6dbd7150506f8119d | | Qwen2-VL | 27.4 | Not recorded | 6a17cbf6dbd7150506f8119d | | LLaVA-N-Inter | 24.2 | Not recorded | 6a17cbf6dbd7150506f8119d | ### ROVER-LSW Narration Score: 50.6. Result ID: 6a17faccfbc373592c11a8e0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8.5B","Input Frames":"core frames","Evaluation Protocol":"Similarity-based"}. This record does not assert independent reproduction. ### ROVER-LSW Narration Score: 49.4. Result ID: 6a17faccfbc373592c11a8e1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8.5B","Input Frames":"core frames","Evaluation Protocol":"LLM-as-a-judge verification"}. This record does not assert independent reproduction. ### VideoEspresso Narration Score: 45.2. Result ID: 6a17faccfbc373592c11a8df. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"8.5B","Input Frames":"core frames"}. This record does not assert independent reproduction. ### LongVA-DPO Narration Score: 35.5. Result ID: 6a17faccfbc373592c11a8dc. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"7.9B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### Qwen-VL-Max Narration Score: 33.9. Result ID: 6a17faccfbc373592c11a8d7. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"–","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### InternVL2 Narration Score: 33.9. Result ID: 6a17faccfbc373592c11a8d9. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.1B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### GPT-4o Narration Score: 32.3. Result ID: 6a17faccfbc373592c11a8d6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Params":"–","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### LLaVA-1.5 Narration Score: 32.3. Result ID: 6a17faccfbc373592c11a8d8. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"6.8B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### LLaVA-N-Video Narration Score: 31.2. Result ID: 6a17faccfbc373592c11a8de. [Reporting paper]() Recorded evaluation configuration: {"Params":"7.1B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### mPLUG-Owl3 Narration Score: 30.6. Result ID: 6a17faccfbc373592c11a8dd. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.1B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### Qwen2-VL Narration Score: 27.4. Result ID: 6a17faccfbc373592c11a8db. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.3B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. ### LLaVA-N-Inter Narration Score: 24.2. Result ID: 6a17faccfbc373592c11a8da. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Params":"8.1B","Input Frames":"full video inputs"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.