Benchmark
SAEBench
What is SAEBench?
SAEBench is a comprehensive benchmark designed to evaluate and compare the performance of language model sparse autoencoders (SAEs). This benchmark provides…
- Released
- 2025-03-12
- Evaluates
- 理解, Understanding, 大语言模型, LLM, 语言理解, Comprehension, 不支持, Unsupported
- Openness
- open
- Importer
- OpenCompass Hub
- Reported scores
- 0
Source provenance
- Original evidence https://hub.opencompass.org.cn/dataset-detail/SAEBench
SAEBench paper, code and dataset
- Paper https://arxiv.org/abs/2402.17700
- Paper https://arxiv.org/abs/2409.14507
- Paper https://arxiv.org/abs/2410.13928
- Paper https://arxiv.org/abs/2410.19278
- Paper https://arxiv.org/abs/2411.18895
- Code https://github.com/adamkarvonen/SAEBench
- Dataset https://huggingface.co/datasets/adamkarvonen/sae_bench_results_0125
No reported scores are on record for this benchmark yet.