wayfind/metask-jev-4b-policy-mix
Upload eval/figs/fig7_scatter.png with huggingface_hub
Upload eval/figs/fig6_board_style.png with huggingface_hub
Upload eval/figs/fig5_latency.png with huggingface_hub
Upload README.md with huggingface_hub
Upload eval/figs/fig7_scatter.png with huggingface_hub
Upload eval/figs/fig6_board_style.png with huggingface_hub
Upload README.md with huggingface_hub
Upload eval/figs/fig7_scatter.png with huggingface_hub
Upload eval/figs/fig6_board_style.png with huggingface_hub
Upload README.md with huggingface_hub
Upload eval/figs/fig8_laya_compare.png with huggingface_hub
Upload eval/figs/fig7_scatter.png with huggingface_hub
Upload eval/figs/fig6_board_style.png with huggingface_hub
Upload eval/figs/fig4_calibration.png with huggingface_hub
Upload eval/figs/fig3_evolution.png with huggingface_hub
Upload eval/figs/fig1_subsets.png with huggingface_hub
Upload README.md with huggingface_hub
Upload temperature.json with huggingface_hub
metask-jev-4b multilingual: +14 locales (MASSIVE), 16-locale checkpoint
selftest: English output
card usability: install+benchmark flow, source notes
fix: judge tier filename mapping
add selftest.sh
card: selftest.sh one-liner
card: JevBench harness local run section (option A/B)
card: copy-paste-ready curl example
add serve.sh
add serve.py
card: HTTP serve section
install.sh: transformers>=4.53 floor
add install.sh to repo
fix: HF repo id to wayfind org
card: native context 262144, 4096 = validated eval point
card: Quickstart with raw Transformers usage
card: one-command install section
card: JevBench board moved to front
card + benchmark figures
weights
config + tokenizer + card
initial commit