frontierphysics
frontierphysics-experiment
FrontierPhysics experiment artifacts
Public artifacts for PR #437, tested at 6989e0ba72c0beb13fc8924c4f9e2461eab57237.
Complete three-trial ZIP, 1.75 GB — anonymously downloaded and SHA-256/CRC verified.
Final report
Raw trial files
All three GPT-6 Astra xhigh trials passed 16/16 tests. Native final rewards: 0.898305, 0, 0.932203. GPT-5.6 Sol xhigh failed one reporting blocker for trial 2; the report flags inconsistent treatment of that criterion across the reviews. Original… See the full description on the dataset page: https://huggingface.co/datasets/benchflow/frontierphysics-experiment.frontierphysics-pr343-trajectoriesfrontierphysics-multiplexing-ion-chain-qnet
Multiplexing ARTIQ experimental results
This dataset is a public snapshot of the HDF5 result files from
artiq_working_dir - 20240315/results, collected for trapped-ion multiplexing
and quantum-networking experiments.
Contents
artiq-results.tar.zst: 7,433 HDF5 result files organized by date and hour.
MANIFEST.sha256: SHA-256 digest for every extracted HDF5 file.
DATASET_INFO.json: source inventory and publication audit summary.
The data span 170 date directories… See the full description on the dataset page: https://huggingface.co/datasets/benchflow/frontierphysics-multiplexing-ion-chain-qnet.frontierphysics-nseos-data
Mirror of two public NICER posterior-sample releases (unmodified)
Byte-identical mirror, used only as a build-time fallback when Zenodo is unavailable.
Primary sources (Zenodo, CC BY 4.0):
nicer/J0030_2spot_RM.txt — PSR J0030+0451, Miller, Lamb, Dittmann et al. 2019 (ApJL 887, L24). Zenodo 10.5281/zenodo.3473466. sha256 747a5cb77be8ea55da5eedf3ff41daa501d1fac426f3fd3d41e6d11088ac9dd2
nicer/NICER+XMM_J0740_RM.txt — PSR J0740+6620, Miller, Lamb, Dittmann et al. 2021 (ApJL 918… See the full description on the dataset page: https://huggingface.co/datasets/solarfury11/frontierphysics-nseos-data.frontierphysics-pr160-trajectories
frontierphysics-pr160-trajectories
Agent-trial run bundle for FrontierPhysics PR #160 (consistent-two-phase-ls-vof).
One folder per graded no-skill rollout (1 codex gpt-5.6-sol, 3 claude claude-opus-5), each with the
complete ACP trajectory (acp_trajectory.jsonl), harness metadata (result.json / timing.json /
config.json / prompts.json), and the verifier outputs (CTRF, pytest stdout, and the artifacts the
verifier collected from the agent workspace). The oracle's artifacts and… See the full description on the dataset page: https://huggingface.co/datasets/ruohai0925/frontierphysics-pr160-trajectories.frontierphysics-pr245-trajectories
FrontierPhysics PR #245 run artifacts
Task: LWIR-image-restoration
Run-artifact bundle (not task inputs). Inputs: https://huggingface.co/datasets/benchflow/heat-assisted-detection-and-ranging
pr245-trajectories.zip: per-run ctrf.json / reward.txt / acp_trajectory.jsonl (#109 / #183 style)
manifest.json
Arm
Pass
oracle
1/1
gpt-5.6-sol no-skill
0/10
gpt-5.6-sol with-skill
1/1
Counted matrix only. Leftover gpt-6-astra / quota / v12 runs are not in this zip.