# Top-k recovery on LLM Judge Comparison full-data Source: [Top-k recovery on LLM Judge Comparison full-data]() Collection index: [llms.txt]() Data updated: 2026-07-03T07:52:02.344Z Evaluation of Top-k recovery performance using recall metrics (naive recall, bias-aware recall) on the full dataset comparison. 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 | Top-k recovery | | Dataset (including any stated split) | LLM Judge Comparison full-data | | Metric | Naive Recall | | 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 - [Naive Recall](): Higher is better; unit: PERCENT. - [Bias-Aware Recall](): Higher is better; unit: PERCENT. - [Delta](): Higher is better; unit: PERCENT. ## Recorded results | Method | Naive Recall | Result date | Report paper | | --- | --- | --- | --- | | DeepSeek-V4-flash | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | DeepSeek-V4-pro | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Claude-Sonnet-4.6 | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | GPT-4.1 | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | GPT-5.5 | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Claude-Opus-4.8 | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Gemini-2.5-pro | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | GPT-5.1 | 100 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Claude-Haiku-4.5 | 93 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Phi-4 | 88 | Not recorded | 6a4741fb143ebdcacacc5b2e | | DeepSeek-chat | 84 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Qwen-2.5 (7B) | 66 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Gemini-2.5-flash-lite | 60 | Not recorded | 6a4741fb143ebdcacacc5b2e | | GPT-4o-mini | 56 | Not recorded | 6a4741fb143ebdcacacc5b2e | | Llama-3.1-8B | 50 | Not recorded | 6a4741fb143ebdcacacc5b2e | | DeepSeek-Coder-6.7B | 8 | Not recorded | 6a4741fb143ebdcacacc5b2e | ### DeepSeek-V4-flash Naive Recall: 100. Result ID: 6a476a2060b88bb339933543. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.37","κˆ (position)":"−0.08"}. This record does not assert independent reproduction. ### DeepSeek-V4-pro Naive Recall: 100. Result ID: 6a476a2060b88bb339933544. [Reporting paper]() Recorded evaluation configuration: {"n":"2","cˆ (verbosity)":"+0.28","κˆ (position)":"+0.01"}. This record does not assert independent reproduction. ### Claude-Sonnet-4.6 Naive Recall: 100. Result ID: 6a476a2060b88bb339933545. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.28","κˆ (position)":"+1.42"}. This record does not assert independent reproduction. ### GPT-4.1 Naive Recall: 100. Result ID: 6a476a2060b88bb339933546. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.24","κˆ (position)":"+1.18"}. This record does not assert independent reproduction. ### GPT-5.5 Naive Recall: 100. Result ID: 6a476a2060b88bb339933548. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.16","κˆ (position)":"+0.07"}. This record does not assert independent reproduction. ### Claude-Opus-4.8 Naive Recall: 100. Result ID: 6a476a2060b88bb339933549. [Reporting paper]() Recorded evaluation configuration: {"n":"2","cˆ (verbosity)":"+0.16","κˆ (position)":"+1.37"}. This record does not assert independent reproduction. ### Gemini-2.5-pro Naive Recall: 100. Result ID: 6a476a2060b88bb33993354a. [Reporting paper]() Recorded evaluation configuration: {"n":"1","cˆ (verbosity)":"+0.15","κˆ (position)":"−0.31"}. This record does not assert independent reproduction. ### GPT-5.1 Naive Recall: 100. Result ID: 6a476a2060b88bb33993354b. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.13","κˆ (position)":"+0.95"}. This record does not assert independent reproduction. ### Claude-Haiku-4.5 Naive Recall: 93. Result ID: 6a476a2060b88bb33993354c. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"−0.16","κˆ (position)":"+0.05"}. This record does not assert independent reproduction. ### Phi-4 Naive Recall: 88. Result ID: 6a476a2060b88bb339933547. [Reporting paper]() Recorded evaluation configuration: {"n":"10","cˆ (verbosity)":"+0.24","κˆ (position)":"+0.32"}. This record does not assert independent reproduction. ### DeepSeek-chat Naive Recall: 84. Result ID: 6a476a2060b88bb339933542. [Reporting paper]() Recorded evaluation configuration: {"n":"5","cˆ (verbosity)":"+0.59","κˆ (position)":"−2.41"}. This record does not assert independent reproduction. ### Qwen-2.5 (7B) Naive Recall: 66. Result ID: 6a476a2060b88bb33993353f. [Reporting paper]() Recorded evaluation configuration: {"n":"10","cˆ (verbosity)":"+0.83","κˆ (position)":"+1.09"}. This record does not assert independent reproduction. ### Gemini-2.5-flash-lite Naive Recall: 60. Result ID: 6a476a2060b88bb339933540. [Reporting paper]() Recorded evaluation configuration: {"n":"3","cˆ (verbosity)":"+0.75","κˆ (position)":"−1.30"}. This record does not assert independent reproduction. ### GPT-4o-mini Naive Recall: 56. Result ID: 6a476a2060b88bb339933541. [Reporting paper]() Recorded evaluation configuration: {"n":"5","cˆ (verbosity)":"+0.61","κˆ (position)":"−0.85"}. This record does not assert independent reproduction. ### Llama-3.1-8B Naive Recall: 50. Result ID: 6a476a2060b88bb33993353e. [Reporting paper]() Recorded evaluation configuration: {"n":"10","cˆ (verbosity)":"+0.99","κˆ (position)":"+0.35"}. This record does not assert independent reproduction. ### DeepSeek-Coder-6.7B Naive Recall: 8. Result ID: 6a476a2060b88bb33993354d. [Reporting paper]() Recorded evaluation configuration: {"n":"5","cˆ (verbosity)":"−0.51","κˆ (position)":"+1.75"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.