{"id":46407330,"url":"https://github.com/decoderesearch/synth-sae-bench-experiments","last_synced_at":"2026-03-05T12:32:58.723Z","repository":{"id":338988198,"uuid":"1156316966","full_name":"decoderesearch/synth-sae-bench-experiments","owner":"decoderesearch","description":"code for the paper: SynthSAEBench: Evaluating Sparse Autoencoders on Scalable Realistic Synthetic Data","archived":false,"fork":false,"pushed_at":"2026-02-17T12:28:24.000Z","size":203,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":0,"default_branch":"main","last_synced_at":"2026-02-17T17:53:38.326Z","etag":null,"topics":["ai","interpretability","machine-learning","sparse-autoencoders"],"latest_commit_sha":null,"homepage":"https://arxiv.org/abs/2602.14687","language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/decoderesearch.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":"CITATION.cff","codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null,"notice":null,"maintainers":null,"copyright":null,"agents":null,"dco":null,"cla":null}},"created_at":"2026-02-12T14:11:37.000Z","updated_at":"2026-02-17T12:28:28.000Z","dependencies_parsed_at":null,"dependency_job_id":null,"html_url":"https://github.com/decoderesearch/synth-sae-bench-experiments","commit_stats":null,"previous_names":["decoderesearch/synth-sae-bench-experiments"],"tags_count":null,"template":false,"template_full_name":null,"purl":"pkg:github/decoderesearch/synth-sae-bench-experiments","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/decoderesearch%2Fsynth-sae-bench-experiments","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/decoderesearch%2Fsynth-sae-bench-experiments/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/decoderesearch%2Fsynth-sae-bench-experiments/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/decoderesearch%2Fsynth-sae-bench-experiments/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/decoderesearch","download_url":"https://codeload.github.com/decoderesearch/synth-sae-bench-experiments/tar.gz/refs/heads/main","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/decoderesearch%2Fsynth-sae-bench-experiments/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":30124493,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-03-05T11:11:57.947Z","status":"ssl_error","status_checked_at":"2026-03-05T11:11:29.001Z","response_time":93,"last_error":"SSL_connect returned=1 errno=0 peeraddr=140.82.121.6:443 state=error: unexpected eof while reading","robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":false,"can_crawl_api":true,"host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["ai","interpretability","machine-learning","sparse-autoencoders"],"created_at":"2026-03-05T12:32:58.617Z","updated_at":"2026-03-05T12:32:58.711Z","avatar_url":"https://github.com/decoderesearch.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# SynthSAEBench\n\nThis repo contains code for the paper: [SynthSAEBench: Evaluating Sparse Autoencoders on Scalable Realistic Synthetic Data](https://arxiv.org/abs/2602.14687). \n\nTrained SAEs and results are online at [https://huggingface.co/decoderesearch/synth-sae-bench-16k-v1-saes](https://huggingface.co/decoderesearch/synth-sae-bench-16k-v1-saes).\n\n## Structure\n\nExperiments for the paper are in the `experiments` directory. The extended SAE classes used in the paper (`XStandardTrainingSAE` and `XJumpReLUTrainingSAE` containing L0 autotuning) are in the `saes` directory.\n\n## Setup\n\nThis project uses uv for package management. To install the dependencies, run:\n\n```bash\nuv sync\n```\n\n## Running Experiments\n\nTo run the experiments, use the `uv run` command. For example, to run the superposition experiment, run:\n\n```bash\nuv run experiments/sweeps/sweep_superposition.py\n```\n\n## Loading the benchmark model\n\nThe main benchmark model is on Huggingface at [decoderesearch/synth-sae-bench-16k-v1](https://huggingface.co/decoderesearch/synth-sae-bench-16k-v1). To load the model, run:\n\n```bash\nfrom sae_lens.synthetic.synthetic_model import SyntheticModel\n\nmodel = SyntheticModel.from_pretrained(\"decoderesearch/synth-sae-bench-16k-v1\")\n```\n\n## Development\n\n### Linting and Formatting\n\nThis project uses ruff for linting and formatting. To run the linting and formatting, run:\n\n```bash\nuv run ruff check .\nuv run ruff format .\n```\n\n### Testing\n\nThis project uses pytest for testing. To run the tests, run:\n\n```bash\nuv run pytest\n```\n\n### Type Checking\n\nThis project uses pyright for type checking. To run the type checking, run:\n\n```bash\nuv run pyright\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdecoderesearch%2Fsynth-sae-bench-experiments","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fdecoderesearch%2Fsynth-sae-bench-experiments","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fdecoderesearch%2Fsynth-sae-bench-experiments/lists"}