# Behavioral Similarity Analysis on τ-Bench and τ2-Bench (test) Source: [Behavioral Similarity Analysis on τ-Bench and τ2-Bench (test)]() Collection index: [llms.txt]() Data updated: 2026-04-24T07:30:07.688Z Evaluation of model performance on Behavioral Similarity Analysis using the test splits of the τ-Bench and τ2-Bench datasets, measured across various structural and linguistic similarity metrics (GED, BERT, N-gram, etc.). 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 | Behavioral Similarity Analysis | | Dataset (including any stated split) | τ-Bench and τ2-Bench (test) | | Metric | GED Score | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 18 of 18 | | 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 - [GED Score](): Higher is better; unit: SCALAR. - [RSE](): Lower is better; unit: SCALAR. - [N-gram Similarity](): Higher is better; unit: PERCENT. - [BERT Similarity](): Higher is better; unit: PERCENT. - [Snode Similarity](): Higher is better; unit: SCALAR. - [Sseq Similarity](): Higher is better; unit: SCALAR. - [Sdep Similarity](): Higher is better; unit: SCALAR. - [AGS Average Similarity](): Higher is better; unit: SCALAR. - [Stylistic Score (Sty.)](): Higher is better; unit: SCALAR. - [Structural Score (Str.)](): Higher is better; unit: SCALAR. - [Alignment Score (Ali.)](): Higher is better; unit: SCALAR. - [RPS Overall Score](): Higher is better; unit: SCALAR. ## Recorded results | Method | GED Score | Result date | Report paper | | --- | --- | --- | --- | | Sonnet 4.5 | 82.6 | Not recorded | 69eaf8f355a3c32af89f8853 | | Opus 4.1 | 79.7 | Not recorded | 69eaf8f355a3c32af89f8853 | | Kimi-K2 | 78.1 | Not recorded | 69eaf8f355a3c32af89f8853 | | Gemini 3 Pro | 77.5 | Not recorded | 69eaf8f355a3c32af89f8853 | | GLM-4.6 | 75.9 | Not recorded | 69eaf8f355a3c32af89f8853 | | GPT-4.1 | 75.2 | Not recorded | 69eaf8f355a3c32af89f8853 | | Kimi-K2 | 74.8 | Not recorded | 69eaf8f355a3c32af89f8853 | | Doubao 1.6 | 72.9 | Not recorded | 69eaf8f355a3c32af89f8853 | | V3.1 | 72.6 | Not recorded | 69eaf8f355a3c32af89f8853 | | Doubao 1.6 | 71.8 | Not recorded | 69eaf8f355a3c32af89f8853 | | R1 | 70.8 | Not recorded | 69eaf8f355a3c32af89f8853 | | Doubao 1.6 | 70.3 | Not recorded | 69eaf8f355a3c32af89f8853 | | Gemini 2.5 Pro | 70.1 | Not recorded | 69eaf8f355a3c32af89f8853 | | V3.1 | 69.2 | Not recorded | 69eaf8f355a3c32af89f8853 | | GPT-5 | 68.3 | Not recorded | 69eaf8f355a3c32af89f8853 | | Qwen3-30B | 65.8 | Not recorded | 69eaf8f355a3c32af89f8853 | | Qwen3-235B | 65.6 | Not recorded | 69eaf8f355a3c32af89f8853 | | V3-0324 | 59.1 | Not recorded | 69eaf8f355a3c32af89f8853 | ### Sonnet 4.5 GED Score: 82.6. Result ID: 69eb1bf547f34ce924d2f397. [Reporting paper]() Recorded evaluation configuration: {"Family":"ANTHROPIC","Thinking Mode":"no-think"}. This record does not assert independent reproduction. ### Opus 4.1 GED Score: 79.7. Result ID: 69eb1bf547f34ce924d2f398. [Reporting paper]() Recorded evaluation configuration: {"Family":"ANTHROPIC","Thinking Mode":"thinking"}. This record does not assert independent reproduction. ### Kimi-K2 GED Score: 78.1. Result ID: 69eb1bf547f34ce924d2f39f. [Reporting paper]() Recorded evaluation configuration: {"Family":"MOONSHOT","Thinking Mode":"thinking"}. This record does not assert independent reproduction. ### Gemini 3 Pro GED Score: 77.5. Result ID: 69eb1bf547f34ce924d2f3a4. [Reporting paper]() Recorded evaluation configuration: {"Family":"GOOGLE"}. This record does not assert independent reproduction. ### GLM-4.6 GED Score: 75.9. Result ID: 69eb1bf547f34ce924d2f3a8. [Reporting paper]() Recorded evaluation configuration: {"Family":"ZHIPU"}. This record does not assert independent reproduction. ### GPT-4.1 GED Score: 75.2. Result ID: 69eb1bf547f34ce924d2f399. [Reporting paper]() Recorded evaluation configuration: {"Family":"OPENAI"}. This record does not assert independent reproduction. ### Kimi-K2 GED Score: 74.8. Result ID: 69eb1bf547f34ce924d2f3a0. [Reporting paper]() Recorded evaluation configuration: {"Family":"MOONSHOT"}. This record does not assert independent reproduction. ### Doubao 1.6 GED Score: 72.9. Result ID: 69eb1bf547f34ce924d2f3a2. [Reporting paper]() Recorded evaluation configuration: {"Family":"BYTEDANCE","Level":"medium"}. This record does not assert independent reproduction. ### V3.1 GED Score: 72.6. Result ID: 69eb1bf547f34ce924d2f39c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Family":"DEEPSEEK","Thinking Mode":"thinking"}. This record does not assert independent reproduction. ### Doubao 1.6 GED Score: 71.8. Result ID: 69eb1bf547f34ce924d2f3a3. [Reporting paper]() Recorded evaluation configuration: {"Family":"BYTEDANCE","Level":"low"}. This record does not assert independent reproduction. ### R1 GED Score: 70.8. Result ID: 69eb1bf547f34ce924d2f39b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Family":"DEEPSEEK"}. This record does not assert independent reproduction. ### Doubao 1.6 GED Score: 70.3. Result ID: 69eb1bf547f34ce924d2f3a1. [Reporting paper]() Recorded evaluation configuration: {"Family":"BYTEDANCE","Level":"high"}. This record does not assert independent reproduction. ### Gemini 2.5 Pro GED Score: 70.1. Result ID: 69eb1bf547f34ce924d2f3a5. [Reporting paper]() Recorded evaluation configuration: {"Family":"GOOGLE"}. This record does not assert independent reproduction. ### V3.1 GED Score: 69.2. Result ID: 69eb1bf547f34ce924d2f39d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Family":"DEEPSEEK","Thinking Mode":"no-think"}. This record does not assert independent reproduction. ### GPT-5 GED Score: 68.3. Result ID: 69eb1bf547f34ce924d2f39a. [Reporting paper]() Recorded evaluation configuration: {"Family":"OPENAI"}. This record does not assert independent reproduction. ### Qwen3-30B GED Score: 65.8. Result ID: 69eb1bf547f34ce924d2f3a6. [Reporting paper]() Recorded evaluation configuration: {"Family":"QWEN","Thinking Mode":"thinking"}. This record does not assert independent reproduction. ### Qwen3-235B GED Score: 65.6. Result ID: 69eb1bf547f34ce924d2f3a7. [Reporting paper]() Recorded evaluation configuration: {"Family":"QWEN","Thinking Mode":"thinking"}. This record does not assert independent reproduction. ### V3-0324 GED Score: 59.1. Result ID: 69eb1bf547f34ce924d2f39e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Family":"DEEPSEEK"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.