-
README: point at both arms
-
Publish the C1 condition: exact per-task prompts, the layer, and its documentation
-
MLAgentBench protocol setup (improve a provided train.py); retire from-scratch tree
-
identify-contrails: note the ~1% positive-pixel imbalance and the all-empty-mask failure mode
-
Hardware section: strict device-resolution requirement (cuda/mps/cpu), forbid CUDA-only checks
-
Hardware hint: accelerator-agnostic (CUDA or Apple MPS)
-
Dockerfiles: auto-select torch build (CUDA on x86_64, CPU on arm64); override with TORCH_VARIANT
-
Scorers: write numeric-only reward.json (Harbor rejects string fields); diagnostics to stdout
-
README: document self-contained task images (no external base image/repo/token)
-
Self-contained task images (FROM python:3.12-slim, own requirements.txt); remove external base-image and agent-specific coupling
-
README: fully agent-neutral (no agent names/specifics), generic -a <agent> run
-
README: drop AutoDS-specific section; add docker-less (-e singularity) run option for any agent
-
Reframe README: general benchmark for any Harbor agent; AutoDS specifics optional
-
README: present Docker (harbor) and venv as two explicit run options
-
Fix stale README/registry: single dataset, agent-neutral base instructions, data shipped on HF
-
Add original-competition evaluation details to the 4 full-data MLAB base instructions
-
Add discretionary GPU hint to base instructions
-
cifar10 -> full 50k/10k, imdb -> full 25k/25k
-
Add harbor scaffolding (task.toml/Dockerfile/prepare/tests) to the 4 full-data tasks
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Add files using upload-large-folder tool
-
Reset mlab-real for 1-folder-per-task restructure
-
Base instructions from original MLAgentBench repo; fix stale metadata (real data, 9 tasks)
-
Re-stage identify-contrails on enlarged real subset (~150 records)
-
Ship real data: spaceship-titanic-base
-
Ship real data: spaceship-titanic
-
Ship real data: ogbn-arxiv-base
-
Ship real data: ogbn-arxiv
-
Ship real data: imdb-base
-
Ship real data: imdb
-
Ship real data: house-price-base
-
Ship real data: house-price
-
Ship real data: feedback-base
-
Ship real data: feedback
-
Ship real data: cifar10-base
-
Ship real data: cifar10
-
Ship real data: amp-parkinsons-base
-
Ship real data: amp-parkinsons
-
Ship real vision task data (fathomnet + contrails)
-
Permissive .gitignore (private mirror ships data)