Safetensors
English
qwen3_moe
mfzzzzzz commited on
Commit
59a534b
路
verified 路
1 Parent(s): 79d12f9

Refine model card layout

Browse files
Files changed (1) hide show
  1. README.md +7 -2
README.md CHANGED
@@ -14,7 +14,12 @@ base_model:
14
  <img src="assets/calibforge-overview.png" alt="Overview of CalibForge and its evaluation results" width="100%">
15
  </p>
16
 
17
- 馃搫 [Paper](https://arxiv.org/abs/2608.06352) 路 馃捇 [Repository](https://github.com/AweAI-Team/CalibForge) 路 馃 [Dataset](https://huggingface.co/datasets/AweAI-Team/CalibForge) 路 馃 [35B Model](https://huggingface.co/AweAI-Team/CalibForge-35B-A3B)
 
 
 
 
 
18
 
19
  CalibForge-30B-A3B is a terminal-agent model fine-tuned from [Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507) using full-parameter, multi-turn supervised fine-tuning. Its training trajectories were distilled from **5,431** calibrated terminal tasks spanning 16 domains. **CalibForge** uses verified solver outcomes and full trajectories to revise executable terminal tasks toward a solver-relative learnable zone. See the [paper](https://arxiv.org/abs/2608.06352) for the task-construction pipeline.
20
 
@@ -28,7 +33,7 @@ CalibForge-30B-A3B is a terminal-agent model fine-tuned from [Qwen3-30B-A3B-Inst
28
  Terminal-Bench 2.0 and Doc2Repo results are reported as mean 卤 SEM over three runs; SWE-bench Pro is evaluated once. Terminal-Bench 2.0 uses [CalibForge-Eval](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/terminal_bench_v2) with a 500-step limit, a one-hour per-task timeout, and sandboxes capped at 16 CPUs and 32 GB RAM. SWE-bench Pro and Doc2Repo use their official evaluation scaffolds; corresponding AweAgent recipes are available for [SWE-bench Pro](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/swe_bench_pro) and [Doc2Repo through BeyondSWE](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/beyond_swe).
29
 
30
  <p align="center">
31
- <img src="assets/tb2-category-results.png" alt="Terminal-Bench 2.0 tasks solved by category for CalibForge-30B-A3B and its base model" width="280">
32
  </p>
33
 
34
  ## Usage with SGLang
 
14
  <img src="assets/calibforge-overview.png" alt="Overview of CalibForge and its evaluation results" width="100%">
15
  </p>
16
 
17
+ <p align="center">
18
+ 馃搫 <a href="https://arxiv.org/abs/2608.06352">Paper</a> 路
19
+ 馃捇 <a href="https://github.com/AweAI-Team/CalibForge">Repository</a> 路
20
+ 馃 <a href="https://huggingface.co/datasets/AweAI-Team/CalibForge">Dataset</a> 路
21
+ 馃 <a href="https://huggingface.co/AweAI-Team/CalibForge-35B-A3B">35B Model</a>
22
+ </p>
23
 
24
  CalibForge-30B-A3B is a terminal-agent model fine-tuned from [Qwen3-30B-A3B-Instruct-2507](https://huggingface.co/Qwen/Qwen3-30B-A3B-Instruct-2507) using full-parameter, multi-turn supervised fine-tuning. Its training trajectories were distilled from **5,431** calibrated terminal tasks spanning 16 domains. **CalibForge** uses verified solver outcomes and full trajectories to revise executable terminal tasks toward a solver-relative learnable zone. See the [paper](https://arxiv.org/abs/2608.06352) for the task-construction pipeline.
25
 
 
33
  Terminal-Bench 2.0 and Doc2Repo results are reported as mean 卤 SEM over three runs; SWE-bench Pro is evaluated once. Terminal-Bench 2.0 uses [CalibForge-Eval](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/terminal_bench_v2) with a 500-step limit, a one-hour per-task timeout, and sandboxes capped at 16 CPUs and 32 GB RAM. SWE-bench Pro and Doc2Repo use their official evaluation scaffolds; corresponding AweAgent recipes are available for [SWE-bench Pro](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/swe_bench_pro) and [Doc2Repo through BeyondSWE](https://github.com/AweAI-Team/AweAgent/tree/main/recipes/beyond_swe).
34
 
35
  <p align="center">
36
+ <img src="assets/tb2-category-results.png" alt="Terminal-Bench 2.0 tasks solved by category for CalibForge-30B-A3B and its base model" width="400">
37
  </p>
38
 
39
  ## Usage with SGLang