# Soft constrained reward optimization on XSum on Gemma-7B Source: [Soft constrained reward optimization on XSum on Gemma-7B]() Collection index: [llms.txt]() Data updated: 2026-05-15T08:42:43.349Z Evaluation of a model's performance using soft constrained reward optimization techniques applied to the XSum summarization dataset, specifically referencing the Gemma-7B model configuration. 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 | Soft constrained reward optimization | | Dataset (including any stated split) | XSum on Gemma-7B | | Metric | Average Soft Constrained Reward | | Unit | SCALAR | | Direction | Higher is better | | Results loaded | 36 of 36 | | 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 - [Average Soft Constrained Reward](): Higher is better; unit: SCALAR. ## Recorded results | Method | Average Soft Constrained Reward | Result date | Report paper | | --- | --- | --- | --- | | MLP-CSR | 0.147 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.147 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.145 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.144 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.143 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.143 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.143 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.143 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.142 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.142 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.142 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.142 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.142 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.141 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.141 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.14 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.14 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.139 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.139 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.137 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.134 | Not recorded | 6a06a87cdbc836b87d829f83 | | MLP-CSR | 0.123 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.122 | Not recorded | 6a06a87cdbc836b87d829f83 | | CSR | 0.117 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.066 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.044 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.037 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.037 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.037 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.036 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.03 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.021 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.021 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.021 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0.015 | Not recorded | 6a06a87cdbc836b87d829f83 | | Uniform | 0 | Not recorded | 6a06a87cdbc836b87d829f83 | ### MLP-CSR Average Soft Constrained Reward: 0.147. Result ID: 6a06dc7447f34ce924ee908c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.147. Result ID: 6a06dc7447f34ce924ee9099. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.145. Result ID: 6a06dc7447f34ce924ee9098. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.144. Result ID: 6a06dc7447f34ce924ee907c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.143. Result ID: 6a06dc7447f34ce924ee907d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.143. Result ID: 6a06dc7447f34ce924ee9087. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.143. Result ID: 6a06dc7447f34ce924ee908a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.143. Result ID: 6a06dc7447f34ce924ee9095. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.142. Result ID: 6a06dc7447f34ce924ee9081. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.142. Result ID: 6a06dc7447f34ce924ee908b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.142. Result ID: 6a06dc7447f34ce924ee908d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.142. Result ID: 6a06dc7447f34ce924ee9094. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.142. Result ID: 6a06dc7447f34ce924ee9097. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.141. Result ID: 6a06dc7447f34ce924ee9088. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.141. Result ID: 6a06dc7447f34ce924ee9093. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.14. Result ID: 6a06dc7447f34ce924ee9080. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.14. Result ID: 6a06dc7447f34ce924ee9089. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.139. Result ID: 6a06dc7447f34ce924ee907f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.139. Result ID: 6a06dc7447f34ce924ee9096. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.137. Result ID: 6a06dc7447f34ce924ee907b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.134. Result ID: 6a06dc7447f34ce924ee9092. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### MLP-CSR Average Soft Constrained Reward: 0.123. Result ID: 6a06dc7447f34ce924ee907e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.122. Result ID: 6a06dc7447f34ce924ee9086. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### CSR Average Soft Constrained Reward: 0.117. Result ID: 6a06dc7447f34ce924ee907a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.066. Result ID: 6a06dc7447f34ce924ee908f. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.044. Result ID: 6a06dc7447f34ce924ee9091. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.037. Result ID: 6a06dc7447f34ce924ee9079. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.037. Result ID: 6a06dc7447f34ce924ee9084. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.037. Result ID: 6a06dc7447f34ce924ee9090. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.036. Result ID: 6a06dc7447f34ce924ee9085. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"10"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.03. Result ID: 6a06dc7447f34ce924ee9078. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"8"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.021. Result ID: 6a06dc7447f34ce924ee9082. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.021. Result ID: 6a06dc7447f34ce924ee9083. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"50","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.021. Result ID: 6a06dc7447f34ce924ee908e. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"100","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0.015. Result ID: 6a06dc7447f34ce924ee9076. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"3"}. This record does not assert independent reproduction. ### Uniform Average Soft Constrained Reward: 0. Result ID: 6a06dc7447f34ce924ee9077. [Reporting paper]() Recorded evaluation configuration: {"Prompt set size (K)":"30","Per-arm budget (b)":"5"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.