Update README.md

Analytics: remove Non-evaluative, trim findings to 3 points, add slogan

Add Analytics tab: interactive Plotly figures + findings for ICLR/ICML/NeurIPS 2025

Fix API fetch: deduplicate reviewers, skip Authors/Rebuttal, add ICML 2025 fields and overall_recommendation rating

Fix NeurIPS/ICML 2025: force API fetch, add strengths_and_weaknesses field, deduplicate reviews

Raise system1_score floor to 3.0 when dominant pattern is gap-consequence enumeration

Dominant pattern governs label: typical concern not best concern; exclude shallow majority from System 2

Refine checklist rule: gap+consequence pattern → System 1; tighten Checklist Inflation bias definition

Add checklist-to-System1 rule: flat lists with low prioritization/density/criticality now prefer System 1

Overhaul review prompt: new 9-dim schema, dominant-mode rule, supporting spans, bias diagnostics

Remove API key label annotation

Patch bundle ratings from API when all ratings are unknown

Fix integer rating extraction; add Manual Input tab for review/meta-review

Update bundle download repo_id to nuojohnchen/Kahneman4Review

Fix meta-review and decision extraction from both bundle and OpenReview API

Add multi-provider/model support (Anthropic, OpenAI, DeepSeek) and consolidate header row

Wire meta-review tab to rate_metareview(), clean up duplicate code in app.py

Fix: remove scale param from gr.Markdown

Add derived metrics (heuristic_dominance, analytic_strength, mixedness) and feedback panel

Replace prompt with full Kahneman dual-process framework

Refactor info sections: same-row buttons with shared expandable panel

Rename to Kahneman4Review, add project description and accordions

Add meta-review tab, decision field, rebuild bundle with metareview/decision

Switch to openreview-py for API access, handles auth and rate limits

Fix OR token: retry on failure, add debug logs

Fix bundle loading: detect LFS pointer, auto-download via hf_hub

Add OpenReview auth via env vars to avoid rate limits

Expand bundle to 32,171 papers (Re² full + ICLR/ICML/NeurIPS 2025)

Simplify app.py: remove monkey-patch, fix startup hang

Fix: remove .submit() on Dropdown, not supported in Gradio 5

Add compressed bundle (Re² test set, 3.5MB), load from gzip

Fix fetcher: v1/v2 API detection, better error message

Fix fetcher: support Re² JSON local cache + v1/v2 API detection

Fix port: use GRADIO_SERVER_PORT env var, default 7860

Initial commit: Review AI Rater Space

initial commit