# Image Captioning on MS COCO v1.3 (val) Source: [Image Captioning on MS COCO v1.3 (val)]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:38:19.988Z Evaluation of Image Captioning performance on the validation split of the MS COCO v1.3 dataset, assessed using CHAIRi, CHAIRs, and Cover 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 | Image Captioning | | Dataset (including any stated split) | MS COCO v1.3 (val) | | Metric | CHAIRi | | 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 - [CHAIRi](): Higher is better; unit: PERCENT. - [CHAIRs](): Higher is better; unit: PERCENT. - [Cover](): Higher is better; unit: PERCENT. ## Recorded results | Method | CHAIRi | Result date | Report paper | | --- | --- | --- | --- | | LLaVA-v1.3 7B | 8.1 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B | 7.4 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 7B PMI | 6.7 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B PMI | 6.5 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B Contrastive | 6.5 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B + LURE | 6.4 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 7B Contrastive | 6.3 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 7B M3ID | 5.9 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 7B M3ID + DPO | 5.7 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B M3ID | 5.5 | Not recorded | 68fa729b1e209131cea5779f | | mPLUG-Owl + LURE | 5.4 | Not recorded | 68fa729b1e209131cea5779f | | LLaVA-v1.3 13B M3ID + DPO | 5.3 | Not recorded | 68fa729b1e209131cea5779f | ### LLaVA-v1.3 7B CHAIRi: 8.1. Result ID: 699262b03a6c3148490e9545. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-7B","Decoding Strategy":"Multinomial"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B CHAIRi: 7.4. Result ID: 699262b03a6c3148490e9549. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Decoding Strategy":"Multinomial"}. This record does not assert independent reproduction. ### LLaVA-v1.3 7B PMI CHAIRi: 6.7. Result ID: 699262b03a6c3148490e9546. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-7B","Decoding Strategy":"PMI"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B PMI CHAIRi: 6.5. Result ID: 699262b03a6c3148490e954a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Decoding Strategy":"PMI"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B Contrastive CHAIRi: 6.5. Result ID: 699262b03a6c3148490e954b. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Decoding Strategy":"Contrastive Decoding"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B + LURE CHAIRi: 6.4. Result ID: 699262b03a6c3148490e954d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Training Method":"LURE"}. This record does not assert independent reproduction. ### LLaVA-v1.3 7B Contrastive CHAIRi: 6.3. Result ID: 699262b03a6c3148490e9547. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-7B","Decoding Strategy":"Contrastive Decoding"}. This record does not assert independent reproduction. ### LLaVA-v1.3 7B M3ID CHAIRi: 5.9. Result ID: 699262b03a6c3148490e9548. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-7B","Decoding Strategy":"M3ID"}. This record does not assert independent reproduction. ### LLaVA-v1.3 7B M3ID + DPO CHAIRi: 5.7. Result ID: 699262b03a6c3148490e954f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-7B","Decoding Strategy":"M3ID","Training Method":"DPO"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B M3ID CHAIRi: 5.5. Result ID: 699262b03a6c3148490e954c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Decoding Strategy":"M3ID"}. This record does not assert independent reproduction. ### mPLUG-Owl + LURE CHAIRi: 5.4. Result ID: 699262b03a6c3148490e954e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"mPLUG-Owl","Training Method":"LURE"}. This record does not assert independent reproduction. ### LLaVA-v1.3 13B M3ID + DPO CHAIRi: 5.3. Result ID: 699262b03a6c3148490e9550. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"LLaVA-v1.3-13B","Decoding Strategy":"M3ID","Training Method":"DPO"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.