nielsr HF Staff commited on
Commit
251dc9e
·
verified ·
1 Parent(s): ba32f8c

Add Terminal-Bench evaluation results

Browse files

This PR adds evaluation results in the [`.eval_results/` format](https://huggingface.co/docs/hub/main/en/eval-results#adding-evaluation-results), so they show up on the official benchmark leaderboards on the Hub.

- **Terminal-Bench 4.0** (`harborframework/terminal-bench`, `terminalbench_4`): 41.9, from the [Artificial Analysis leaderboard](https://artificialanalysis.ai/evaluations/terminalbench-4-0) (GLM-5.3 (Max)).
- **Terminal-Bench-Science 0.1** (`harborframework/terminal-bench-science`, `terminalbenchscience_0_1`): 9.5, from the [Artificial Analysis leaderboard](https://artificialanalysis.ai/evaluations/terminal-bench-science) (GLM-5.3 (Max)).

Artificial Analysis runs these benchmarks independently with the mini-swe-agent harness and reports pass@1 averaged over three repeats per task.

.eval_results/terminal-bench-4.0.yaml ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ - dataset:
2
+ id: harborframework/terminal-bench
3
+ task_id: terminalbench_4
4
+ value: 41.9
5
+ source:
6
+ url: https://artificialanalysis.ai/evaluations/terminalbench-4-0
7
+ name: "Artificial Analysis Terminal-Bench 4.0 leaderboard"
8
+ user: ArtificialAnalysis
9
+ notes: "GLM-5.3 (Max); mini-swe-agent harness, pass@1 averaged over three repeats."
.eval_results/terminal-bench-science.yaml ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ - dataset:
2
+ id: harborframework/terminal-bench-science
3
+ task_id: terminalbenchscience_0_1
4
+ value: 9.5
5
+ source:
6
+ url: https://artificialanalysis.ai/evaluations/terminal-bench-science
7
+ name: "Artificial Analysis Terminal-Bench-Science leaderboard"
8
+ user: ArtificialAnalysis
9
+ notes: "GLM-5.3 (Max); mini-swe-agent harness, pass@1 averaged over three repeats."