# Shortcut detection on BenchMarker Shortcuts Source: [Shortcut detection on BenchMarker Shortcuts]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:30:50.967Z Evaluation of model performance on detecting shortcuts within the BenchMarker Shortcuts dataset, measured using Accuracy, F1 Score, and Cohen's Kappa. 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 | Shortcut detection | | Dataset (including any stated split) | BenchMarker Shortcuts | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 26 of 26 | | 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. - [F1 Score](): Higher is better; unit: PERCENT. - [Cohen's Kappa](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | GPT-5 | 81.6 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Claude 4.5 | 80.8 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Claude 4.5 | 78.2 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | GPT-5 | 76.8 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemini 2.5 | 75.7 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Command R | 74.2 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemma-3 | 74.2 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 73.4 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Command R+ | 72.3 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 70.5 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemini 2.5 | 70.1 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemini 2.5 | 68.6 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | GPT-5 | 67.9 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Always Not Flawed | 63.5 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 62.4 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 61.3 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemma-3 | 61.3 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | LLaMA-3.1 | 60.5 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | LLaMA-3.2 | 57.9 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 57.2 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | LLaMA-3.2 | 56.8 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Gemma-3 | 50.2 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Random (50/50) | 50 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | LLaMA-3.1 | 48.7 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Qwen-3 | 42.1 | Not recorded | 69896a28c97a0b3e3b5dc5ac | | Always Flawed | 36.5 | Not recorded | 69896a28c97a0b3e3b5dc5ac | ### GPT-5 Accuracy: 81.6. Result ID: 699646735698665ec48a55ff. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Claude 4.5 Accuracy: 80.8. Result ID: 699646735698665ec48a5601. [Reporting paper]() Recorded evaluation configuration: {"Model Variant":"Sonnet"}. This record does not assert independent reproduction. ### Claude 4.5 Accuracy: 78.2. Result ID: 699646735698665ec48a5600. [Reporting paper]() Recorded evaluation configuration: {"Model Variant":"Haiku"}. This record does not assert independent reproduction. ### GPT-5 Accuracy: 76.8. Result ID: 699646735698665ec48a55fe. [Reporting paper]() Recorded evaluation configuration: {"Model Variant":"Mini"}. This record does not assert independent reproduction. ### Gemini 2.5 Accuracy: 75.7. Result ID: 699646735698665ec48a55fa. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Variant":"Lite"}. This record does not assert independent reproduction. ### Command R Accuracy: 74.2. Result ID: 699646735698665ec48a5602. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Gemma-3 Accuracy: 74.2. Result ID: 699646735698665ec48a560c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"27B"}. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 73.4. Result ID: 699646735698665ec48a5609. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"32B"}. This record does not assert independent reproduction. ### Command R+ Accuracy: 72.3. Result ID: 699646735698665ec48a5603. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 70.5. Result ID: 699646735698665ec48a5604. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"0.6B"}. This record does not assert independent reproduction. ### Gemini 2.5 Accuracy: 70.1. Result ID: 699646735698665ec48a55fc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Variant":"Pro"}. This record does not assert independent reproduction. ### Gemini 2.5 Accuracy: 68.6. Result ID: 699646735698665ec48a55fb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Variant":"Flash"}. This record does not assert independent reproduction. ### GPT-5 Accuracy: 67.9. Result ID: 699646735698665ec48a55fd. [Reporting paper]() Recorded evaluation configuration: {"Model Variant":"Nano"}. This record does not assert independent reproduction. ### Always Not Flawed Accuracy: 63.5. Result ID: 699646735698665ec48a5612. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 62.4. Result ID: 699646735698665ec48a5605. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"1.7B"}. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 61.3. Result ID: 699646735698665ec48a5608. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"14B"}. This record does not assert independent reproduction. ### Gemma-3 Accuracy: 61.3. Result ID: 699646735698665ec48a560b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"12B"}. This record does not assert independent reproduction. ### LLaMA-3.1 Accuracy: 60.5. Result ID: 699646735698665ec48a5610. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"70B"}. This record does not assert independent reproduction. ### LLaMA-3.2 Accuracy: 57.9. Result ID: 699646735698665ec48a560e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"3B"}. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 57.2. Result ID: 699646735698665ec48a5607. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"8B"}. This record does not assert independent reproduction. ### LLaMA-3.2 Accuracy: 56.8. Result ID: 699646735698665ec48a560d. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"1B"}. This record does not assert independent reproduction. ### Gemma-3 Accuracy: 50.2. Result ID: 699646735698665ec48a560a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"4B"}. This record does not assert independent reproduction. ### Random (50/50) Accuracy: 50. Result ID: 699646735698665ec48a5611. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LLaMA-3.1 Accuracy: 48.7. Result ID: 699646735698665ec48a560f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Parameters":"8B"}. This record does not assert independent reproduction. ### Qwen-3 Accuracy: 42.1. Result ID: 699646735698665ec48a5606. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Parameters":"4B"}. This record does not assert independent reproduction. ### Always Flawed Accuracy: 36.5. Result ID: 699646735698665ec48a5613. [Reporting paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.