# Instruction Tuning on Alpaca instruction-tuning 52k Source: [Instruction Tuning on Alpaca instruction-tuning 52k]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:57:15.745Z Evaluation of model performance on Instruction Tuning using the Alpaca instruction-tuning 52k dataset, measured by Pairwise Winning Score. 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 | Instruction Tuning | | Dataset (including any stated split) | Alpaca instruction-tuning 52k | | Metric | Pairwise Winning Score | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 19 of 19 | | 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 - [Pairwise Winning Score](): Higher is better; unit: PERCENT. ## Recorded results | Method | Pairwise Winning Score | Result date | Report paper | | --- | --- | --- | --- | | GRADFILTERING | 116 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 116 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 115 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 113 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 112 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Superfiltering | 110 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 110 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 107 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 106 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 106 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Superfiltering | 102 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Superfiltering | 102 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 101 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Random Split | 97 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Random Split | 93 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | Random Split | 92 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 1.2 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 1.19 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | | GRADFILTERING | 1.1 | Not recorded | 69705dc0e25ad2fa9d1ed4a5 | ### GRADFILTERING Pairwise Winning Score: 116. Result ID: 6996798e5698665ec4c12881. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"Full","Training data fraction":"5%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 116. Result ID: 6996798e5698665ec4c12882. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"Full","Training data fraction":"10%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 115. Result ID: 6996798e5698665ec4c12887. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"15%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 113. Result ID: 6996798e5698665ec4c1288a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-13B","Adaptation regime":"Full","Training data fraction":"10%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 112. Result ID: 6996798e5698665ec4c12883. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"Full","Training data fraction":"15%"}. This record does not assert independent reproduction. ### Superfiltering Pairwise Winning Score: 110. Result ID: 6996798e5698665ec4c1287c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"10%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 110. Result ID: 6996798e5698665ec4c12889. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-13B","Adaptation regime":"LoRA","Training data fraction":"15%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 107. Result ID: 6996798e5698665ec4c1287f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"10%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 106. Result ID: 6996798e5698665ec4c12880. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"15%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 106. Result ID: 6996798e5698665ec4c12888. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"Full","Training data fraction":"10%"}. This record does not assert independent reproduction. ### Superfiltering Pairwise Winning Score: 102. Result ID: 6996798e5698665ec4c1287b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"5%"}. This record does not assert independent reproduction. ### Superfiltering Pairwise Winning Score: 102. Result ID: 6996798e5698665ec4c1287d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"15%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 101. Result ID: 6996798e5698665ec4c1287e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"5%"}. This record does not assert independent reproduction. ### Random Split Pairwise Winning Score: 97. Result ID: 6996798e5698665ec4c12878. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"5%"}. This record does not assert independent reproduction. ### Random Split Pairwise Winning Score: 93. Result ID: 6996798e5698665ec4c12879. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"10%"}. This record does not assert independent reproduction. ### Random Split Pairwise Winning Score: 92. Result ID: 6996798e5698665ec4c1287a. [Reporting paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-7B","Adaptation regime":"LoRA","Training data fraction":"15%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 1.2. Result ID: 6996798e5698665ec4c12885. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-13B","Adaptation regime":"LoRA","Training data fraction":"10%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 1.19. Result ID: 6996798e5698665ec4c12884. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-13B","Adaptation regime":"LoRA","Training data fraction":"5%"}. This record does not assert independent reproduction. ### GRADFILTERING Pairwise Winning Score: 1.1. Result ID: 6996798e5698665ec4c12886. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Base Model":"LLaMA2-13B","Adaptation regime":"Full","Training data fraction":"10%"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.