# Feature Attribution on FEVER Source: [Feature Attribution on FEVER]() Collection index: [llms.txt]() Data updated: 2026-06-02T11:47:44.991Z Evaluation of Feature Attribution methods on the FEVER dataset, measured using Comprehensiveness and Sufficiency metrics. 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 | Feature Attribution | | Dataset (including any stated split) | FEVER | | Metric | Comprehensiveness | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 33 of 33 | | 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 - [Comprehensiveness](): Higher is better; unit: SCALAR. - [Sufficiency](): Higher is better; unit: SCALAR. ## Recorded results | Method | Comprehensiveness | Result date | Report paper | | --- | --- | --- | --- | | AttnLRP | 0.75 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.75 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.69 | Not recorded | 68fa7bbf1e209131cea8a2be | | AttnLRP | 0.68 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.68 | Not recorded | 68fa7bbf1e209131cea8a2be | | AttnLRP | 0.65 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.62 | Not recorded | 68fa7bbf1e209131cea8a2be | | AttnLRP | 0.61 | Not recorded | 68fa7bbf1e209131cea8a2be | | AttnLRP | 0.6 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.54 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.53 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.53 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.51 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.51 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.5 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.47 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.47 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.47 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.46 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.43 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.43 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.42 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.39 | Not recorded | 68fa7bbf1e209131cea8a2be | | GIM | 0.39 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.33 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.32 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.31 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.28 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.2 | Not recorded | 68fa7bbf1e209131cea8a2be | | DeepLIFT | 0.12 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.06 | Not recorded | 68fa7bbf1e209131cea8a2be | | GxI | 0.03 | Not recorded | 68fa7bbf1e209131cea8a2be | | IG | 0.03 | Not recorded | 68fa7bbf1e209131cea8a2be | ### AttnLRP Comprehensiveness: 0.75. Result ID: 6a1ec2d3fbc373592c1a2865. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.75. Result ID: 6a1ec2d3fbc373592c1a286c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.69. Result ID: 6a1ec2d3fbc373592c1a286d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"8B"}. This record does not assert independent reproduction. ### AttnLRP Comprehensiveness: 0.68. Result ID: 6a1ec2d3fbc373592c1a2868. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.68. Result ID: 6a1ec2d3fbc373592c1a286f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"3B"}. This record does not assert independent reproduction. ### AttnLRP Comprehensiveness: 0.65. Result ID: 6a1ec2d3fbc373592c1a2866. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"8B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.62. Result ID: 6a1ec2d3fbc373592c1a286b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"1B"}. This record does not assert independent reproduction. ### AttnLRP Comprehensiveness: 0.61. Result ID: 6a1ec2d3fbc373592c1a2867. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"1.5B"}. This record does not assert independent reproduction. ### AttnLRP Comprehensiveness: 0.6. Result ID: 6a1ec2d3fbc373592c1a2864. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"1B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.54. Result ID: 6a1ec2d3fbc373592c1a285c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.53. Result ID: 6a1ec2d3fbc373592c1a2852. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"3B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.53. Result ID: 6a1ec2d3fbc373592c1a2861. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"8B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.51. Result ID: 6a1ec2d3fbc373592c1a2859. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.51. Result ID: 6a1ec2d3fbc373592c1a286e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"1.5B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.5. Result ID: 6a1ec2d3fbc373592c1a285a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"8B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.47. Result ID: 6a1ec2d3fbc373592c1a2855. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"3B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.47. Result ID: 6a1ec2d3fbc373592c1a2856. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"2B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.47. Result ID: 6a1ec2d3fbc373592c1a2860. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.46. Result ID: 6a1ec2d3fbc373592c1a2853. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"8B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.43. Result ID: 6a1ec2d3fbc373592c1a2854. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"1.5B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.43. Result ID: 6a1ec2d3fbc373592c1a2863. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"3B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.42. Result ID: 6a1ec2d3fbc373592c1a2869. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"2B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.39. Result ID: 6a1ec2d3fbc373592c1a2851. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"1B"}. This record does not assert independent reproduction. ### GIM Comprehensiveness: 0.39. Result ID: 6a1ec2d3fbc373592c1a286a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"9B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.33. Result ID: 6a1ec2d3fbc373592c1a285f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"1B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.32. Result ID: 6a1ec2d3fbc373592c1a2857. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"9B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.31. Result ID: 6a1ec2d3fbc373592c1a2862. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"1.5B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.28. Result ID: 6a1ec2d3fbc373592c1a2858. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"LLAMA","Parameter Count":"1B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.2. Result ID: 6a1ec2d3fbc373592c1a285e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"9B"}. This record does not assert independent reproduction. ### DeepLIFT Comprehensiveness: 0.12. Result ID: 6a1ec2d3fbc373592c1a285d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"2B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.06. Result ID: 6a1ec2d3fbc373592c1a2850. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"9B"}. This record does not assert independent reproduction. ### GxI Comprehensiveness: 0.03. Result ID: 6a1ec2d3fbc373592c1a284f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Gemma","Parameter Count":"2B"}. This record does not assert independent reproduction. ### IG Comprehensiveness: 0.03. Result ID: 6a1ec2d3fbc373592c1a285b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen","Parameter Count":"1.5B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.