# Reward Modeling on RewardBench 1k Source: [Reward Modeling on RewardBench 1k]() Collection index: [llms.txt]() Data updated: 2026-04-10T07:20:53.278Z Evaluation of model performance on the RewardBench 1k dataset for the Reward Modeling task, measured by Positional Consistency. 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 | Reward Modeling | | Dataset (including any stated split) | RewardBench 1k | | Metric | Positional Consistency | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | 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 - [Positional Consistency](): Higher is better; unit: PERCENT. ## Recorded results | Method | Positional Consistency | Result date | Report paper | | --- | --- | --- | --- | | ReflectRM | 84.9 | Not recorded | 69d883e07d0172140ec77804 | | Unified Training | 83.6 | Not recorded | 69d883e07d0172140ec77804 | | ReflectRM | 81.1 | Not recorded | 69d883e07d0172140ec77804 | | RFT | 78.8 | Not recorded | 69d883e07d0172140ec77804 | | Unified Training | 78.5 | Not recorded | 69d883e07d0172140ec77804 | | Qwen3-8B | 78.4 | Not recorded | 69d883e07d0172140ec77804 | | RFT | 75.1 | Not recorded | 69d883e07d0172140ec77804 | | Qwen3-4B | 74.8 | Not recorded | 69d883e07d0172140ec77804 | | ReflectRM | 68.3 | Not recorded | 69d883e07d0172140ec77804 | | RFT | 67 | Not recorded | 69d883e07d0172140ec77804 | | Unified Training | 64.3 | Not recorded | 69d883e07d0172140ec77804 | | Qwen3-8B | 60.9 | Not recorded | 69d883e07d0172140ec77804 | | ReflectRM | 59.4 | Not recorded | 69d883e07d0172140ec77804 | | Unified Training | 58.4 | Not recorded | 69d883e07d0172140ec77804 | | RFT | 56.5 | Not recorded | 69d883e07d0172140ec77804 | | Qwen3-4B | 52.3 | Not recorded | 69d883e07d0172140ec77804 | ### ReflectRM Positional Consistency: 84.9. Result ID: 69d8a4ba47f34ce924c11a87. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"Unified Training","Inference Strategy":"Two-Stage Inference"}. This record does not assert independent reproduction. ### Unified Training Positional Consistency: 83.6. Result ID: 69d8a4ba47f34ce924c11a86. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"Unified Training"}. This record does not assert independent reproduction. ### ReflectRM Positional Consistency: 81.1. Result ID: 69d8a4ba47f34ce924c11a83. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"Unified Training","Inference Strategy":"Two-Stage Inference"}. This record does not assert independent reproduction. ### RFT Positional Consistency: 78.8. Result ID: 69d8a4ba47f34ce924c11a85. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"RFT"}. This record does not assert independent reproduction. ### Unified Training Positional Consistency: 78.5. Result ID: 69d8a4ba47f34ce924c11a82. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"Unified Training"}. This record does not assert independent reproduction. ### Qwen3-8B Positional Consistency: 78.4. Result ID: 69d8a4ba47f34ce924c11a84. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B"}. This record does not assert independent reproduction. ### RFT Positional Consistency: 75.1. Result ID: 69d8a4ba47f34ce924c11a81. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"RFT"}. This record does not assert independent reproduction. ### Qwen3-4B Positional Consistency: 74.8. Result ID: 69d8a4ba47f34ce924c11a80. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B"}. This record does not assert independent reproduction. ### ReflectRM Positional Consistency: 68.3. Result ID: 69d8a4ba47f34ce924c11a93. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"Unified Training","Inference Strategy":"Two-Stage Inference"}. This record does not assert independent reproduction. ### RFT Positional Consistency: 67. Result ID: 69d8a4ba47f34ce924c11a95. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"RFT"}. This record does not assert independent reproduction. ### Unified Training Positional Consistency: 64.3. Result ID: 69d8a4ba47f34ce924c11a92. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"Unified Training"}. This record does not assert independent reproduction. ### Qwen3-8B Positional Consistency: 60.9. Result ID: 69d8a4ba47f34ce924c11a94. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B"}. This record does not assert independent reproduction. ### ReflectRM Positional Consistency: 59.4. Result ID: 69d8a4ba47f34ce924c11a97. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"Unified Training","Inference Strategy":"Two-Stage Inference"}. This record does not assert independent reproduction. ### Unified Training Positional Consistency: 58.4. Result ID: 69d8a4ba47f34ce924c11a96. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-8B","Training Method":"Unified Training"}. This record does not assert independent reproduction. ### RFT Positional Consistency: 56.5. Result ID: 69d8a4ba47f34ce924c11a91. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B","Training Method":"RFT"}. This record does not assert independent reproduction. ### Qwen3-4B Positional Consistency: 52.3. Result ID: 69d8a4ba47f34ce924c11a90. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen3-4B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.