Config: exact copy of evalscope dpv4-int8_nothinking.yaml (generation_config + dataset_args + agent_config, zero custom fields)

Co-Authored-By: Claude <noreply@anthropic.com>

evalharness/config/README.md

@@ -26,7 +26,6 @@ generation:
 run:
   repeats: 12          # 跑 12 遍取均值（temp=1 方差测量用）
   limit: null          # 全量 / limit_per_task: 10 每子集 10 条
-  concurrency: 4
   checkpoint: true
   resume: true
 

evalharness/config/dp4-nothink.yaml

@@ -1,88 +1,193 @@
-default:
-  temperature: 0.0
-  max_tokens: 32768
-  top_p: 1.0
 gpqa_diamond:
-  temperature: 1.0
-  max_tokens: 8192
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 hle:
-  limit_per_task: 25
-  judge: dp4-flash
-  judge_url: http://174.1.51.4:30000/v1
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 aime24:
-  temperature: 1.0
-  repeats: 12
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 aime25:
-  temperature: 1.0
-  repeats: 12
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 mmlu_pro:
-  max_tokens: 8192
-  limit_per_task: 20
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 simple_qa:
-  max_tokens: 8192
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 arc:
-  max_tokens: 8192
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 bbh:
-  limit_per_task: 10
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 live_code_bench:
-  temperature: 1.0
-  limit: 200
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
+  dataset_args:
+    subset_list:
+    - release_v6
 aime26:
-  temperature: 1.0
-  repeats: 12
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 hmmt26:
-  temperature: 1.0
-  repeats: 12
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 imo_answerbench:
-  temperature: 1.0
-  limit_per_task: 25
-  judge: dp4-flash
-  judge_url: http://174.1.51.4:30000/v1
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 drop:
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 hellaswag:
-  max_tokens: 8192
-  limit: 400
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 mmlu:
-  max_tokens: 8192
-  limit_per_task: 4
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 openai_mrcr:
-  max_tokens: 8192
-  limit: 200
-  max_input_tokens: 120000
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 bigcodebench:
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 humaneval:
-  temperature: 1.0
+  generation_config:
+    temperature: 1.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 gsm8k:
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 competition_math:
-  limit_per_task: 40
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768
 cmmlu:
-  max_tokens: 8192
-  limit_per_task: 3
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 trivia_qa:
-  max_tokens: 8192
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 winogrande:
-  max_tokens: 8192
-  limit: 200
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
 longbench_v2:
-  max_tokens: 8192
-  limit_per_task: 66
-  max_input_tokens: 120000
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 8192
+  dataset_args:
+    subset_list:
+    - short
+    - medium
+    - long
 tau2_bench:
-  max_tokens: 16384
-  env: tau2_official
-  max_turns: 50
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 16384
+  dataset_args:
+    extra_params:
+      user_model: deepseek-v4-pro
+      api_key: sk-9ed86ef546ca47e3afa7c3b014dea268
+      api_base: https://api.deepseek.com/v1
+      generation_config:
+        temperature: 0.0
+        max_tokens: 4096
+  agent_config:
+    mode: native
+    strategy: react
+    max_steps: 50
 general_fc:
-  max_tokens: 4096
-  limit: 400
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 4096
 bfcl_v3:
-  max_tokens: 4096
-  limit_per_task: 20
-  env: bfcl_mock
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 4096
+    parallel_tool_calls: true
+  dataset_args:
+    extra_params:
+      is_fc_model: true
+      underscore_to_dot: true
 swe_bench_verified:
-  limit: 70
+  generation_config:
+    temperature: 0.0
+    top_p: 1.0
+    stream: true
+    max_tokens: 32768