# Language Modeling on Shakespeare 1M params (val) Source: [Language Modeling on Shakespeare 1M params (val)]() Collection index: [llms.txt]() Data updated: 2026-05-25T07:27:10.408Z Evaluation of Language Modeling performance on the validation split of the 1 Million parameter version of the Shakespeare dataset, measured using Negative Log-Likelihood (NLL). 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 | Language Modeling | | Dataset (including any stated split) | Shakespeare 1M params (val) | | Metric | NLL | | Unit | SCALAR | | Direction | Lower is better | | Results loaded | 35 of 35 | | 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 - [NLL](): Lower is better; unit: SCALAR. ## Recorded results | Method | NLL | Result date | Report paper | | --- | --- | --- | --- | | Tensor Cache | 4.79 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 4.8 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 4.91 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 4.91 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 4.91 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 4.96 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 5.14 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 5.24 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 5.32 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 5.47 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 5.48 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 5.51 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 5.6 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 5.66 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 5.7 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 5.73 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 5.79 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 5.79 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 5.82 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 5.87 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 5.92 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Tensor Cache | 5.98 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 6.02 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Infini-attention | 6.06 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 6.09 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 6.1 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Full KV | 6.13 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 6.16 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 6.16 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 6.21 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 6.24 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 6.24 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 6.25 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | StreamingLLM | 6.25 | Not recorded | 6a13d7816f71c5ae7dc9e814 | | Window KV | 6.27 | Not recorded | 6a13d7816f71c5ae7dc9e814 | ### Tensor Cache NLL: 4.79. Result ID: 6a13f8dffbc373592c0abfc1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"256"}. This record does not assert independent reproduction. ### Infini-attention NLL: 4.8. Result ID: 6a13f8dffbc373592c0abfba. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"256"}. This record does not assert independent reproduction. ### Full KV NLL: 4.91. Result ID: 6a13f8dffbc373592c0abfa5. [Reporting paper]() Recorded evaluation configuration: {"context length L":"256"}. This record does not assert independent reproduction. ### Window KV NLL: 4.91. Result ID: 6a13f8dffbc373592c0abfac. [Reporting paper]() Recorded evaluation configuration: {"context length L":"256"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 4.91. Result ID: 6a13f8dffbc373592c0abfb3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"256"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 4.96. Result ID: 6a13f8dffbc373592c0abfc3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"1024"}. This record does not assert independent reproduction. ### Infini-attention NLL: 5.14. Result ID: 6a13f8dffbc373592c0abfbc. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"1024"}. This record does not assert independent reproduction. ### Full KV NLL: 5.24. Result ID: 6a13f8dffbc373592c0abfa7. [Reporting paper]() Recorded evaluation configuration: {"context length L":"1024"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 5.32. Result ID: 6a13f8dffbc373592c0abfc7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"16384"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 5.47. Result ID: 6a13f8dffbc373592c0abfb5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"1024"}. This record does not assert independent reproduction. ### Window KV NLL: 5.48. Result ID: 6a13f8dffbc373592c0abfae. [Reporting paper]() Recorded evaluation configuration: {"context length L":"1024"}. This record does not assert independent reproduction. ### Infini-attention NLL: 5.51. Result ID: 6a13f8dffbc373592c0abfc0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"16384"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 5.6. Result ID: 6a13f8dffbc373592c0abfc4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"2048"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 5.66. Result ID: 6a13f8dffbc373592c0abfc5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"4096"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 5.7. Result ID: 6a13f8dffbc373592c0abfc6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"8192"}. This record does not assert independent reproduction. ### Full KV NLL: 5.73. Result ID: 6a13f8dffbc373592c0abfab. [Reporting paper]() Recorded evaluation configuration: {"context length L":"16384"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 5.79. Result ID: 6a13f8dffbc373592c0abfb9. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"16384"}. This record does not assert independent reproduction. ### Infini-attention NLL: 5.79. Result ID: 6a13f8dffbc373592c0abfbd. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"2048"}. This record does not assert independent reproduction. ### Window KV NLL: 5.82. Result ID: 6a13f8dffbc373592c0abfb2. [Reporting paper]() Recorded evaluation configuration: {"context length L":"16384"}. This record does not assert independent reproduction. ### Infini-attention NLL: 5.87. Result ID: 6a13f8dffbc373592c0abfbe. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"4096"}. This record does not assert independent reproduction. ### Infini-attention NLL: 5.92. Result ID: 6a13f8dffbc373592c0abfbf. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"8192"}. This record does not assert independent reproduction. ### Tensor Cache NLL: 5.98. Result ID: 6a13f8dffbc373592c0abfc2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"512"}. This record does not assert independent reproduction. ### Full KV NLL: 6.02. Result ID: 6a13f8dffbc373592c0abfa8. [Reporting paper]() Recorded evaluation configuration: {"context length L":"2048"}. This record does not assert independent reproduction. ### Infini-attention NLL: 6.06. Result ID: 6a13f8dffbc373592c0abfbb. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"512"}. This record does not assert independent reproduction. ### Full KV NLL: 6.09. Result ID: 6a13f8dffbc373592c0abfa9. [Reporting paper]() Recorded evaluation configuration: {"context length L":"4096"}. This record does not assert independent reproduction. ### Full KV NLL: 6.1. Result ID: 6a13f8dffbc373592c0abfa6. [Reporting paper]() Recorded evaluation configuration: {"context length L":"512"}. This record does not assert independent reproduction. ### Full KV NLL: 6.13. Result ID: 6a13f8dffbc373592c0abfaa. [Reporting paper]() Recorded evaluation configuration: {"context length L":"8192"}. This record does not assert independent reproduction. ### Window KV NLL: 6.16. Result ID: 6a13f8dffbc373592c0abfaf. [Reporting paper]() Recorded evaluation configuration: {"context length L":"2048"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 6.16. Result ID: 6a13f8dffbc373592c0abfb6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"2048"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 6.21. Result ID: 6a13f8dffbc373592c0abfb8. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"8192"}. This record does not assert independent reproduction. ### Window KV NLL: 6.24. Result ID: 6a13f8dffbc373592c0abfb1. [Reporting paper]() Recorded evaluation configuration: {"context length L":"8192"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 6.24. Result ID: 6a13f8dffbc373592c0abfb7. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"4096"}. This record does not assert independent reproduction. ### Window KV NLL: 6.25. Result ID: 6a13f8dffbc373592c0abfad. [Reporting paper]() Recorded evaluation configuration: {"context length L":"512"}. This record does not assert independent reproduction. ### StreamingLLM NLL: 6.25. Result ID: 6a13f8dffbc373592c0abfb4. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"context length L":"512"}. This record does not assert independent reproduction. ### Window KV NLL: 6.27. Result ID: 6a13f8dffbc373592c0abfb0. [Reporting paper]() Recorded evaluation configuration: {"context length L":"4096"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.