Datasetpublic

dougalldeepmind/2026-09-10-strongreject-qwen3-6-27b-lora-9284-numina-control-716-r64

strongreject eval of matboz/qwen3.6-27b-lora-9284-numina-control-716-r64 (mode=think) field value experiment strongreject eval of matboz/qwen3.6-27b-lora-9284-numina-control-716-r64 (mode=think) date_generated 2026-09-10 constitution none source_repo teaching_claude_why_replication @ 8f5659b77847773963bfac26319efb4319bd6f6f models {"target": "matboz/qwen3.6-27b-lora-9284-numina-control-716-r64", "target_revision": "edfb4287c10f553c541ba28216f202d0c0f47055"… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-09-10-strongreject-qwen3-6-27b-lora-9284-numina-control-716-r64.

sourceHugging Faceupdated 17d agoView on Hugging Face

0likes64downloads

Dataset Card

strongreject eval of matboz/qwen3.6-27b-lora-9284-numina-control-716-r64 (mode=think)

field	value
`experiment`	strongreject eval of matboz/qwen3.6-27b-lora-9284-numina-control-716-r64 (mode=think)
`date_generated`	2026-09-10
`constitution`	none
`source_repo`	teachingclaudewhy_replication @ 8f5659b77847773963bfac26319efb4319bd6f6f
`models`	{"target": "matboz/qwen3.6-27b-lora-9284-numina-control-716-r64", "targetrevision": "edfb4287c10f553c541ba28216f202d0c0f47055", "base": "Qwen/Qwen3.6-27B", "baserevision": "6a9e13bd6fc8f0983b9b99948120bc37f49c13e9"}
`generation_config`	{"temperature": 1.0, "topp": 0.95, "maxtokens": 16384, "requesttimeout": 600, "maxretries": 2}
`schema`	rollouts/: self-contained transcripts; results/: results.json + judge/eval outputs; metadata/: run_meta.json + config + provenance
`provenance`	/home/matthewb/git repos/teachingclaudewhyreplication/.venv/bin/evals --name strongreject --target matboz/qwen3.6-27b-lora-9284-numina-control-716-r64 --server root@64.247.201.53:19269 nprompts=313 njailbreaks=20 passes=5 worstofnthreshold=0.5 judge.model=google/gemini-3-flash-preview concurrency.generation=24 concurrency.judge=24 --terminate-pod