Update README.md
Browse files
README.md
CHANGED
|
@@ -32,7 +32,3 @@ The dataset was originally used to train a 4B parameter model via pure SFT (no R
|
|
| 32 |
- Batch size: 128, 10 training epochs
|
| 33 |
- Training framework: verl
|
| 34 |
|
| 35 |
-
## Limitations
|
| 36 |
-
- Synthetic dataset generated via LLM-based synthesis with executor validation; may contain rare fallback binding artifacts
|
| 37 |
-
- Optimized for multi-turn tool-use tasks; not intended for general chat or reasoning benchmarks
|
| 38 |
-
- Tool coverage is limited to the function sets from BFCL v4 and τ²-bench environments
|
|
|
|
| 32 |
- Batch size: 128, 10 training epochs
|
| 33 |
- Training framework: verl
|
| 34 |
|
|
|
|
|
|
|
|
|
|
|
|