qwen3_6
Datasets
All datasets matching “qwen3_6”2026-09-11-dh-qwen3-6-27b-lora-9284-numina-control-716-r64
Delegated-harm evaluation with corrected scoring of saved rollouts
field
value
experiment
Delegated-harm evaluation with corrected scoring of saved rollouts
date_generated
2026-09-11
constitution
none
source_repo
teaching_claude_why_replication @ d627d0587a2980069b7700e72f727dae594c9f49
models
{"hf_path": "matboz/qwen3.6-27b-lora-9284-numina-control-716-r64", "base_model": "Qwen/Qwen3.6-27B", "adapter": true, "mode": "think", "model_key":… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-09-11-dh-qwen3-6-27b-lora-9284-numina-control-716-r64.qwen36-27b-length-traces
Qwen3.6-27B generation-length prediction: heads, calibrations and workloads
Artifacts for conformal length-aware LLM scheduling on Qwen/Qwen3.6-27B — predicting a
request's remaining generation length from a hidden layer during decoding, wrapping it in a
split-conformal interval, and scheduling with SRPT inside vLLM. Extends TRAIL
(Don't Stop Me Now, ICLR'25) to a hybrid-attention reasoning model.
This repo contains the derived artifacts, not the raw activations. The 3250… See the full description on the dataset page: https://huggingface.co/datasets/dungnv/qwen36-27b-length-traces.2026-09-11-dh-qwen36-lora-table2-9284-difficult-advice-chunk-only-702-rank-64-dynbatch
Delegated-harm evaluation with corrected scoring of saved rollouts
field
value
experiment
Delegated-harm evaluation with corrected scoring of saved rollouts
date_generated
2026-09-11
constitution
none
source_repo
teaching_claude_why_replication @ d627d0587a2980069b7700e72f727dae594c9f49
models
{"hf_path": "dougalldeepmind/2026-08-21-qwen36-lora-table2-9284-difficult-advice-chunk-only-702-rank-64-dynbatch", "base_model": "Qwen/Qwen3.6-27B", "adapter": true… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-09-11-dh-qwen36-lora-table2-9284-difficult-advice-chunk-only-702-rank-64-dynbatch.2026-09-12-dh-qwen36-lora-table2-9284-nonmoral-deliberation-684-rank-64-dynbatch
Delegated-harm evaluation with corrected scoring of saved rollouts
field
value
experiment
Delegated-harm evaluation with corrected scoring of saved rollouts
date_generated
2026-09-12
constitution
none
source_repo
teaching_claude_why_replication @ 7cb72cf09f15fb58839862e1ef1c6930562b3cca
models
{"hf_path": "dougalldeepmind/2026-09-02-qwen36-lora-table2-9284-nonmoral-deliberation-684-rank-64-dynbatch", "base_model": "Qwen/Qwen3.6-27B", "adapter": true, "mode":… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-09-12-dh-qwen36-lora-table2-9284-nonmoral-deliberation-684-rank-64-dynbatch.2026-09-23-swebench-qwen36-0-nosynth
Full 300-task SWE-bench Lite no-DA control; partial until valid coverage and grading finish
field
value
experiment
Full 300-task SWE-bench Lite no-DA control; partial until valid coverage and grading finish
date_generated
2026-09-23
constitution
none
source_repo
teaching_claude_why_replication; exact sources and SHA256 in metadata/source
models
dougalldeepmind/2026-09-22-qwen36-0-nosynth@633908b72a9799fb3e6b101b0a8a82aec3c3d642; base… See the full description on the dataset page: https://huggingface.co/datasets/dougalldeepmind/2026-09-23-swebench-qwen36-0-nosynth.qwen36-arithmetic-readouts
Arithmetic Intermediate Readout Sensitivity on Qwen3.6-27B
Arithmetic intermediate detection with a Jacobian lens depends sharply on the prompt token being read and the numeral forms accepted by the scorer. Across 105 order-of-operations items, the recorded hosted-lens responses contain the intermediate at rank 1 on 48 items at the trailing space, versus 4 at the preceding token. On the 25 held-out items with two-digit intermediates, rank-1 detection falls from 13 to 3 when… See the full description on the dataset page: https://huggingface.co/datasets/ec75hash/qwen36-arithmetic-readouts.