Fetching metadata from the HF Docker repository...
-
Goal-binding default prompt + temp 0.2 (9/9 traps now catch); think-split in Trap-It-Yourself; accurate judge keywords; honest context claim
-
UI/UX redesign: tap-to-play trap card grid, animated verdict banner, stat-tile scoreboard, fixed hero/tab contrast
-
Submission metadata: track:wood, sponsor:nvidia, off-brand/off-grid/sharing badges
-
Parse <think> reasoning: show final answer, collapse reasoning, judge on answer
-
Rebuild as Gotcha Arena: reasoning-trap deck + scoreboard + Blind-Spot Check
-
Restore easy benchmark examples
-
Remove easy examples
-
Add goal-binding examples and 1k token default
-
Serve GGUF with persistent llama.cpp server
-
Run GGUF through official llama.cpp CLI
-
Build latest llama.cpp binding from source
-
Upgrade llama.cpp binding for Nemotron GGUF
-
Match Spaces runtime dependency
-
Enable real llama.cpp GGUF inference
-
Keep benchmark examples visible
-
Add benchmark-style prompt examples
-
Polish First-Principle AI interface
-
Point app to org model repo
-
Add First-Principle AI Gradio app
-
initial commit
