flashvenom commited on
Commit
e0499fe
·
verified ·
1 Parent(s): c1b42be

Upload folder using huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +14 -1
README.md CHANGED
@@ -156,9 +156,22 @@ The guarantees live in the decoding harness, so inference goes through the repo:
156
  git clone https://github.com/nikshepsvn/thimble
157
  cd thimble && uv venv && uv pip install -e .
158
  # put thimble-v6.pt in checkpoints/, tokenizer.json in data/
159
- python scripts/final_eval.py --ckpt thimble-v6 --suite seal-tools-in
 
 
 
 
 
 
 
 
 
 
160
  ```
161
 
 
 
 
162
  ## Integrity
163
 
164
  Public corpora (xlam, ToolACE, Dolci, Glaive, official benchmark train splits)
 
156
  git clone https://github.com/nikshepsvn/thimble
157
  cd thimble && uv venv && uv pip install -e .
158
  # put thimble-v6.pt in checkpoints/, tokenizer.json in data/
159
+
160
+ python demo.py "make a reservation at Nobu for 2 people at 7pm and text Sam saying dinner is on"
161
+ # [{"name": "createReservation",
162
+ # "arguments": {"partySize": 2, "restaurant": "Nobu", "time": "7pm"}},
163
+ # {"name": "sendMessage",
164
+ # "arguments": {"body": "dinner is on", "contact": "Sam"}}]
165
+
166
+ python demo.py "sing me a happy birthday song"
167
+ # [] (refused: no tool applies)
168
+
169
+ python scripts/final_eval.py --ckpt thimble-v6 --suite seal-tools-in # reproduce the table
170
  ```
171
 
172
+ Real output, not a mock — typed integers, two-call composition, and refusal,
173
+ with structure guaranteed by the grammar.
174
+
175
  ## Integrity
176
 
177
  Public corpora (xlam, ToolACE, Dolci, Glaive, official benchmark train splits)