# Question Answering on GPQA Diamond (Accuracy, Tokens, Token Reduction) Source: [Question Answering on GPQA Diamond (Accuracy, Tokens, Token Reduction)]() Collection index: [llms.txt]() Data updated: 2026-07-01T08:01:56.516Z Evaluation of model performance on the GPQA Diamond dataset for Question Answering, focusing on accuracy, token count, and the percentage reduction in tokens used. 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 | Question Answering | | Dataset (including any stated split) | GPQA Diamond | | Metric | Accuracy | | 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 - [Accuracy](): Higher is better; unit: PERCENT. - [Tokens](): Lower is better; unit: SCALAR. - [Token Reduction (%)](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Fork-think | 70.1 | Not recorded | 6a449f12e0224bd4f49930df | | Parallel Thinking | 67.3 | Not recorded | 6a449f12e0224bd4f49930df | | Parallel Thinking | 66.9 | Not recorded | 6a449f12e0224bd4f49930df | | Fork-think | 63.2 | Not recorded | 6a449f12e0224bd4f49930df | | Parallel Thinking | 62.1 | Not recorded | 6a449f12e0224bd4f49930df | | Fork-think | 62.1 | Not recorded | 6a449f12e0224bd4f49930df | | Greedy | 61.1 | Not recorded | 6a449f12e0224bd4f49930df | | CoT | 60.8 | Not recorded | 6a449f12e0224bd4f49930df | | CoT | 55.2 | Not recorded | 6a449f12e0224bd4f49930df | | CoT | 53.9 | Not recorded | 6a449f12e0224bd4f49930df | | Greedy | 39.8 | Not recorded | 6a449f12e0224bd4f49930df | | Greedy | 31.3 | Not recorded | 6a449f12e0224bd4f49930df | ### Fork-think Accuracy: 70.1. Result ID: 6a44c97373d097d845bef7d6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi-4-RP-14B","Decoding Strategy":"Fork-think","Branching factor (n)":"32","Seed path length (l0)":"2,048"}. This record does not assert independent reproduction. ### Parallel Thinking Accuracy: 67.3. Result ID: 6a44c97373d097d845bef7d5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi-4-RP-14B","Decoding Strategy":"Parallel Thinking","Branching factor (n)":"32"}. This record does not assert independent reproduction. ### Parallel Thinking Accuracy: 66.9. Result ID: 6a44c97373d097d845bef7d1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepSeek-8B","Decoding Strategy":"Parallel Thinking","Branching factor (n)":"32"}. This record does not assert independent reproduction. ### Fork-think Accuracy: 63.2. Result ID: 6a44c97373d097d845bef7ce. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-8B","Decoding Strategy":"Fork-think","Branching factor (n)":"32","Seed path length (l0)":"2,048"}. This record does not assert independent reproduction. ### Parallel Thinking Accuracy: 62.1. Result ID: 6a44c97373d097d845bef7cd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-8B","Decoding Strategy":"Parallel Thinking","Branching factor (n)":"32"}. This record does not assert independent reproduction. ### Fork-think Accuracy: 62.1. Result ID: 6a44c97373d097d845bef7d2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepSeek-8B","Decoding Strategy":"Fork-think","Branching factor (n)":"32","Seed path length (l0)":"2,048"}. This record does not assert independent reproduction. ### Greedy Accuracy: 61.1. Result ID: 6a44c97373d097d845bef7cf. [Reporting paper]() Recorded evaluation configuration: {"Model":"DeepSeek-8B","Decoding Strategy":"Greedy"}. This record does not assert independent reproduction. ### CoT Accuracy: 60.8. Result ID: 6a44c97373d097d845bef7d0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepSeek-8B","Decoding Strategy":"CoT"}. This record does not assert independent reproduction. ### CoT Accuracy: 55.2. Result ID: 6a44c97373d097d845bef7cc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen3-8B","Decoding Strategy":"CoT"}. This record does not assert independent reproduction. ### CoT Accuracy: 53.9. Result ID: 6a44c97373d097d845bef7d4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Phi-4-RP-14B","Decoding Strategy":"CoT"}. This record does not assert independent reproduction. ### Greedy Accuracy: 39.8. Result ID: 6a44c97373d097d845bef7cb. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen3-8B","Decoding Strategy":"Greedy"}. This record does not assert independent reproduction. ### Greedy Accuracy: 31.3. Result ID: 6a44c97373d097d845bef7d3. [Reporting paper]() Recorded evaluation configuration: {"Model":"Phi-4-RP-14B","Decoding Strategy":"Greedy"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.