# Referring Expression Comprehension on RefCOCO (Accuracy, Average Tokens) Source: [Referring Expression Comprehension on RefCOCO (Accuracy, Average Tokens)]() Collection index: [llms.txt]() Data updated: 2026-06-09T10:44:20.800Z Evaluation of Referring Expression Comprehension performance on the RefCOCO dataset, reporting Accuracy and Average Tokens. 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 | Referring Expression Comprehension | | Dataset (including any stated split) | RefCOCO | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 7 of 7 | | 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](): Higher is better; unit: PERCENT. - [Average Tokens](): Lower is better; unit: SCALAR. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Visual Para-Thinker++ | 88.1 | Not recorded | 6a279e5b90966ade919bfa4c | | Visual Para-Thinker++ | 84.4 | Not recorded | 6a279e5b90966ade919bfa4c | | Base MLLM + Multi-Agent Debate | 83 | Not recorded | 6a279e5b90966ade919bfa4c | | Base MLLM + Para-Thinker | 82.6 | Not recorded | 6a279e5b90966ade919bfa4c | | Base MLLM + Self-Consistency | 81.7 | Not recorded | 6a279e5b90966ade919bfa4c | | Base MLLM + Long CoT | 80.8 | Not recorded | 6a279e5b90966ade919bfa4c | | Base MLLM | 79.7 | Not recorded | 6a279e5b90966ade919bfa4c | ### Visual Para-Thinker++ Accuracy: 88.1. Result ID: 6a27ed5cfbc373592c25725b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Training Stage (MAC/RDMAO)":"MAC+RDMAO"}. This record does not assert independent reproduction. ### Visual Para-Thinker++ Accuracy: 84.4. Result ID: 6a27ed5cfbc373592c25725a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Training Stage (MAC/RDMAO)":"MAC only"}. This record does not assert independent reproduction. ### Base MLLM + Multi-Agent Debate Accuracy: 83. Result ID: 6a27ed5cfbc373592c257259. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### Base MLLM + Para-Thinker Accuracy: 82.6. Result ID: 6a27ed5cfbc373592c257258. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","K (Number of Paths/Agents)":"4"}. This record does not assert independent reproduction. ### Base MLLM + Self-Consistency Accuracy: 81.7. Result ID: 6a27ed5cfbc373592c257257. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","K (Number of Paths/Agents)":"4"}. This record does not assert independent reproduction. ### Base MLLM + Long CoT Accuracy: 80.8. Result ID: 6a27ed5cfbc373592c257256. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### Base MLLM Accuracy: 79.7. Result ID: 6a27ed5cfbc373592c257255. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Decoding Strategy":"greedy"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.