Add TestRouter leaderboard submission

[TreeTXT]

Summary

• add the TestRouter configuration and full/robustness prediction files
• register qwen/qwen3.7-plus as a universal model and add its token pricing
• add SinapisAI inference support for the four-model TestRouter pool

Model pool

• qwen/qwen3.5-flash-02-23
• deepseek/deepseek-v4-flash
• qwen/qwen3.7-plus
• deepseek/deepseek-v4-pro

The Qwen3.7 Plus prices are expressed in USD per million tokens using the supplied conversion rate of 1 USD = 6.74 CNY.

Local evaluation

• RouterArena score: 0.7036
• accuracy: 70.68%
• total cost: $1.237440
• average cost per 1K queries: $0.1473
• robustness score: 0.9143
• optimal selection: 0.7605
• optimal cost: 0.6439
• optimal accuracy: 0.8293

Validation

• python router_inference/check_config_prediction_files.py TestRouter full --check-generated-result
• python router_inference/check_config_prediction_files.py TestRouter robustness
• python -m py_compile llm_inference/model_inference.py universal_model_names.py
• git diff --check

Both RouterArena validation runs passed. The full file contains 8,400 regular entries plus 2,427 optimality entries, and the robustness file contains 420 entries.

[TreeTXT]

/evaluate

[TreeTXT]

/evaluate

[github-actions]

Router Evaluation Results

Router: TestRouter
Dataset Split: full

RouterArena Metrics

Metric	Value
RouterArena Score	0.7030
Accuracy	70.62%
Total Cost	$1.237440
Avg Cost per Query	$0.000147
Avg Cost per 1K Queries	$0.1473
Number of Queries	8400
Abnormal Entries	14
Robustness Score	0.9143

⚠️ 14 of 8400 queries (0.2%) had no valid generation (inference failed / empty answer) and were scored as incorrect (0). These queries still count toward the denominator, so accuracy and cost reflect the full query set. Please regenerate predictions for these queries and resubmit for a complete evaluation.

Optimality Metrics

Metric	Value
Opt.Sel (Optimal Selection)	0.7590
Opt.Cost (Cost Efficiency)	0.6406
Opt.Acc (Accuracy vs Optimal)	0.8278

---

Evaluation completed by RouterArena automated workflow

[github-actions]

Router Evaluation Results

Router: TestRouter
Dataset Split: full

RouterArena Metrics

Metric	Value
RouterArena Score	0.7030
Accuracy	70.62%
Total Cost	$1.237440
Avg Cost per Query	$0.000147
Avg Cost per 1K Queries	$0.1473
Number of Queries	8400
Abnormal Entries	14
Robustness Score	0.9143

⚠️ 14 of 8400 queries (0.2%) had no valid generation (inference failed / empty answer) and were scored as incorrect (0). These queries still count toward the denominator, so accuracy and cost reflect the full query set. Please regenerate predictions for these queries and resubmit for a complete evaluation.

Optimality Metrics

Metric	Value
Opt.Sel (Optimal Selection)	0.7590
Opt.Cost (Cost Efficiency)	0.6406
Opt.Acc (Accuracy vs Optimal)	0.8278

---

Evaluation completed by RouterArena automated workflow