Benchmark-Qwen-Ref: v0.21.12
在将 DSW 包代理范围限定为仅验证器依赖出站流量后，进行完整的端到端重跑：Release -> DSW SWE-bench Verified (500) -> Terminal-Bench 2.0 (89) -> 同一 Release 发布。
<!-- qwen-code-dsw-swe-verified:start -->
### SWE-bench Verified
- 状态：**隔离（QUARANTINED）**
- 数据集：`swe-bench/swe-bench-verified@2`；500/500 已完成
- 结果：293 个已解决，123 个未解决，3 个执行错误，81 个基础设施故障
- 得分：未发布（不可评分或隔离运行）
- 得分分母：有效评分器结果（293 + 123 = 416）
- Qwen Code：`v0.21.12` (`b965d5f8c24f`)；模型 `qwen3.7-plus`
- 运行：`pool-8d03378ba2724d3b`；调度工作流：https://github.com/QwenLM/qwen-code/actions/runs/31922258423
- 案例轨迹：`swe-bench-verified-dsw-eas-full-20260816-r3-trajectories.tar.gz`（421 个案例）
<!-- qwen-code-dsw-swe-verified:end -->
<!-- qwen-code-dsw-terminal-bench-2.0:start -->
### Terminal-Bench 2.0
- 状态：**隔离（QUARANTINED）**
- 数据集：`
[email protected]`；89/89 已完成
- 结果：1 个已解决，14 个未解决，29 个执行错误，45 个基础设施故障
- 得分：未发布（不可评分或隔离运行）
- 得分分母：有效评分器结果（1 + 14 = 15）
- Qwen Code：`v0.21.12` (`b965d5f8c24f`)；模型 `qwen3.7-plus`
- 运行：`pool-57a2e6a1dc41426d`；调度工作流：https://github.com/QwenLM/qwen-code/actions/runs/31922258423
- 案例轨迹：`terminal-bench-2.0-dsw-eas-full-20260816-r3-trajectories.tar.gz`（17 个案例）
<!-- qwen-code-dsw-terminal-bench-2.0:end -->
Benchmark-Qwen-Ref: v0.21.12
Full end-to-end rerun after scoping the DSW package proxy to verifier dependency egress only: Release -> DSW SWE-bench Verified (500) -> Terminal-Bench 2.0 (89) -> same Release publication.
<!-- qwen-code-dsw-swe-verified:start -->
### SWE-bench Verified
- Status: **QUARANTINED**
- Dataset: `swe-bench/swe-bench-verified@2`; 500/500 completed
- Results: 293 resolved, 123 unresolved, 3 execution errors, 81 infrastructure failures
- Score: not published (non-scoreable or quarantined run)
- Score denominator: valid grader results (293 + 123 = 416)
- Qwen Code: `v0.21.12` (`b965d5f8c24f`); model `qwen3.7-plus`
- Run: `pool-8d03378ba2724d3b`; dispatch workflow: https://github.com/QwenLM/qwen-code/actions/runs/31922258423
- Case trajectories: `swe-bench-verified-dsw-eas-full-20260816-r3-trajectories.tar.gz` (421 cases)
<!-- qwen-code-dsw-swe-verified:end -->
<!-- qwen-code-dsw-terminal-bench-2.0:start -->
### Terminal-Bench 2.0
- Status: **QUARANTINED**
- Dataset: `
[email protected]`; 89/89 completed
- Results: 1 resolved, 14 unresolved, 29 execution errors, 45 infrastructure failures
- Score: not published (non-scoreable or quarantined run)
- Score denominator: valid grader results (1 + 14 = 15)
- Qwen Code: `v0.21.12` (`b965d5f8c24f`); model `qwen3.7-plus`
- Run: `pool-57a2e6a1dc41426d`; dispatch workflow: https://github.com/QwenLM/qwen-code/actions/runs/31922258423
- Case trajectories: `terminal-bench-2.0-dsw-eas-full-20260816-r3-trajectories.tar.gz` (17 cases)
<!-- qwen-code-dsw-terminal-bench-2.0:end -->