README: add paper link (huggingface.co/papers/2605.07075) + collection link

README: add weights-download section + cross-link to corpus v1/v2 dataset repos

Trim metric dropdown and shorten landing copy

Deploy MLPMetricFull v2 (47k models, with ID emb)

Surface modality filter on the UI: family column + status line showing required capability and eligible-candidate count

Add task-modality sanity filter: drop LLMs from image-gen / ASR / VQA results

Expand official-pretrained filter: ~60 orgs + bare-name foundation pattern (StepFun, THUDM, Moonshot, Baichuan, InternLM, ByteDance, etc.)

Shorten BYOK notice; rename Recommend button to Search

Pin starlette<0.40 and fastapi<0.115 to keep gradio 4.44 TemplateResponse positional-args compat

Revert gradio to 4.44.0 to match HF Space sdk_version (avoids dep resolution conflict)

Pin jinja2<3.1.5 to fix gradio 4.44 unhashable dict TypeError; bump gradio to 4.44.1

Pin Python version to 3.11 for prebuilt numpy/pandas/torch wheels

Deploy ModelLens v1: BYOK OpenAI key, size filter, official-only filter, 47k HF model pool

initial commit