Safetensors
qwen3
Zichen1024 commited on
Commit
9defe2c
·
verified ·
1 Parent(s): 9184174

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +3 -3
README.md CHANGED
@@ -30,7 +30,7 @@ SAP-4B is a 4B-parameter agentic language model specialized in **long-horizon mu
30
 
31
  ## Evaluation Results
32
 
33
- ### 1. BFCL v4 Multi-Turn (Latest Re-Evaluation, Updated Setting)
34
 
35
  | Benchmark Subset | Accuracy |
36
  |------------------|----------|
@@ -40,7 +40,7 @@ SAP-4B is a 4B-parameter agentic language model specialized in **long-horizon mu
40
  | multi_turn_miss_func | 24.50% |
41
  | multi_turn_miss_param | 26.00% |
42
 
43
- ### 2. τ²-bench (Paper Submission Version)
44
 
45
  | Benchmark Split | Score |
46
  |-----------------|-------|
@@ -50,7 +50,7 @@ SAP-4B is a 4B-parameter agentic language model specialized in **long-horizon mu
50
 
51
  ---
52
 
53
- ### Performance Note (Discrepancy with Submitted Paper)
54
  - The paper has been submitted to EMNLP 2026, and its reported results use the benchmark evaluation setting available at the time of submission.
55
  - For BFCL v4 Multi-Turn: subsequent updates to the evaluation setting led to minor subset-level score fluctuations in the latest re-evaluation, while the overall average performance remains consistent with the paper's conclusions. Fluctuations are purely caused by evaluation pipeline adjustments (stricter matching rules for base/long-context subsets, more favorable calibration for robustness subsets), with no changes to the model or training data.
56
  - For τ²-bench: only the submission-version result is currently available; re-evaluation under an updated setting may produce slightly different subset-level scores.
 
30
 
31
  ## Evaluation Results
32
 
33
+ ### 1. BFCL v4 Multi-Turn
34
 
35
  | Benchmark Subset | Accuracy |
36
  |------------------|----------|
 
40
  | multi_turn_miss_func | 24.50% |
41
  | multi_turn_miss_param | 26.00% |
42
 
43
+ ### 2. τ²-bench
44
 
45
  | Benchmark Split | Score |
46
  |-----------------|-------|
 
50
 
51
  ---
52
 
53
+ ### Performance Note
54
  - The paper has been submitted to EMNLP 2026, and its reported results use the benchmark evaluation setting available at the time of submission.
55
  - For BFCL v4 Multi-Turn: subsequent updates to the evaluation setting led to minor subset-level score fluctuations in the latest re-evaluation, while the overall average performance remains consistent with the paper's conclusions. Fluctuations are purely caused by evaluation pipeline adjustments (stricter matching rules for base/long-context subsets, more favorable calibration for robustness subsets), with no changes to the model or training data.
56
  - For τ²-bench: only the submission-version result is currently available; re-evaluation under an updated setting may produce slightly different subset-level scores.