update

data/dataset/agent-nxpay-id-40-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69c2687adb5fe6ae6082cc5504270a3b592b01226a43defb38bb0d6eac829206
+size 211683

data/eval_data/siliconflow/siliconflow/Qwen#QwQ-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e5ee4af8e0b636f992510850946aea0c891d718736305adcd318024cd1a8796f
+size 2436409

data/eval_data/siliconflow/siliconflow/Qwen#QwQ-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a657945f6f5b28396c9a538fdbc623a9c20926f518a3f93e29398f6073557923
+size 258605

data/eval_data/siliconflow/siliconflow/Qwen#QwQ-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4f2b83d2a03d1f0fd54c84bb1bbf1407609ff032279e61ebabc3762b3a7db6f5
+size 1211300

data/eval_data/siliconflow/siliconflow/Qwen#QwQ-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb2580363c9d95aae96e632f939076a8d814877176601919170dc44d335f5bce
+size 892580

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-14B/shenzhen_sase/siliconflow_api_key/20250813_174600/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f0fa41a9e6a68d7da60199c8ed27d19ef78f7486262188558270b1380313fa3
+size 2397396

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-14B/shenzhen_sase/siliconflow_api_key/20250813_174600/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92cf0ca215afbc52cd93fc66733a0527ca5a491e20dc317173f901ce6e1c87f7
+size 258545

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-14B/shenzhen_sase/siliconflow_api_key/20250813_174600/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41d37a7948993c79855004ab8d6ba71feb2a4e334b20e0c695d62766addfb764
+size 1210547

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-14B/shenzhen_sase/siliconflow_api_key/20250813_174600/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b63327ba9e8252bb7c2ec9b7f31d7a9e2df4dc2c46e72d58487afab6635a54c
+size 873820

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-14B/shenzhen_sase/siliconflow_api_key/20250814_175318/agent-bingoplus-ph-25-summary.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb5ff498f24964d44ce0417035016816e15cb023c26086ece9e4a05e4442bca1
+size 180171

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-235B-A22B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_115346/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7444f40db5ed93a76208b9021a684e04d75b5a69296b8169818e84f4cef7f635
+size 2409152

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-235B-A22B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_115346/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc3c5ebde1353edf931b29f97467860cad09ee499d51b4d15c2a2c6f7658db8e
+size 258403

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-235B-A22B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_115346/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a7f7f13c1c26217a73818ce6e105850c0c1550a123835bf5c84ad2b6c710e8a
+size 1211617

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-235B-A22B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_115346/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f54b8f450d9cc4b08183ddfc4b2c34a50279568aec5e0f11c212347e73b2c907
+size 879114

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_113756/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a359ea25e35f20f507fab92af54770cc543b36ebc218babf2fbc386a93aba2c6
+size 2405978

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_113756/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:94a6383a6fb0ef83b3c82692b726646eb15d967bfac746ae90124df612b1dbb0
+size 258539

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_113756/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:82826bd4ce3d64e02ce1ea79fc3d73a28654e3f1d51e45b4536754c24c9d9993
+size 1211170

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B-Instruct-2507/shenzhen_sase/siliconflow_api_key/20250814_113756/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ea6308aced7c2ccfa2c7f5de09229887a8990815fb6491c79482d59144b9a4f2
+size 882503

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B/shenzhen_sase/siliconflow_api_key/20250814_093406/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:875912816e6e396e929359a6c5a44521a3b0d51e0d28754d28566edffb540a85
+size 2404865

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B/shenzhen_sase/siliconflow_api_key/20250814_093406/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ff6a056cc471b5af50d00092b76eef9dae03673972ce70eb41b46609eb503ac3
+size 258671

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B/shenzhen_sase/siliconflow_api_key/20250814_093406/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b48b94d7c1389ba21506dea94d6df957e87d1bd70e42939636f2d9e9fba663ed
+size 1211401

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-30B-A3B/shenzhen_sase/siliconflow_api_key/20250814_093406/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4dbb2418dc5e76b73a5fb725a7e2e7e0aff2e9c95f648055a4a83a38cc39138c
+size 873914

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-32B/shenzhen_sase/siliconflow_api_key/20250814_175528/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e47033bde0f0c387f0d9bfae7c9d32cf06aa9dd4e0e62a0225dcad2410dd952c
+size 2414736

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-32B/shenzhen_sase/siliconflow_api_key/20250814_175528/agent-bingoplus-ph-25-summary.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:505a2de4045efa8e27c03c8b1dbe5094e3d522e9b3484073ea469d412d7c0898
+size 182566

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-32B/shenzhen_sase/siliconflow_api_key/20250814_175528/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b665bf4e3e9fa32cf273ddc923c8e04850625d05819462e5987b6fa1cb44b4c8
+size 258583

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-32B/shenzhen_sase/siliconflow_api_key/20250814_175528/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:283f3c29a5a2b9b7d0faae4300bdacd28a930a387db1a8754ca6488af5f39429
+size 1208679

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-32B/shenzhen_sase/siliconflow_api_key/20250814_175528/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90cf5b4f17d7b27c1e20f73aa24fddd2895fcc8a9f1598a015b3d0b384e6c090
+size 881952

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-8B/shenzhen_sase/siliconflow_api_key/20250814_163605/agent-bingoplus-ph-25-summary.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b8c2b975eec172c895c866473d67a6d25a0e6c27eba6f260f5b14c718b5c293
+size 181707

data/eval_data/siliconflow/siliconflow/Qwen#Qwen3-8B/shenzhen_sase/siliconflow_api_key/20250814_163605/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:571d1fc583aa78a42b2d4c5cc7e9bdff71b4e4a472adf03459ef9913aec7bca0
+size 258615

data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-200-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb4c714e099894700646c482bd68c997d41b14acb16255ec1b2795f11ddace61
+size 2409276

data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-90-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ddac1a4f6815eee7123c9fc8fcd4b056ce5a180bb8c56b35c442bd0c19ac62f0
+size 258516

data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-400-choice.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bcb61871b9ebed78a80efec356d8c047d38cfda9049767c082cdb429f67962c0
+size 1211414

data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-80-chat.jsonl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e09e738040b67bd357dcbc85cbf3b5566a8e5597a24921c4a514d3d7e2434f1c
+size 882489

examples/make_dataset/make_bingoplus_ph_200_chat.py

@@ -17,12 +17,14 @@ def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument(
         "--raw_dataset",
-        default=(project_path / "data/raw_dataset/finished/agent-bingoplus-ph-200-chat").as_posix(),
+        # default=(project_path / "data/raw_dataset/finished/agent-bingoplus-ph-200-chat").as_posix(),
+        default=(project_path / "data/raw_dataset/agent-nxpay-id-100-chat").as_posix(),
         type=str
     )
     parser.add_argument(
         "--dataset",
-        default=(project_path / "data/dataset/agent-bingoplus-ph-200-chat.jsonl").as_posix(),
+        # default=(project_path / "data/dataset/agent-bingoplus-ph-200-chat.jsonl").as_posix(),
+        default=(project_path / "data/dataset/agent-nxpay-id-100-chat.jsonl").as_posix(),
         type=str
     )
     args = parser.parse_args()

examples/make_raw_dataset/step_3_filter_by_keywords.py

@@ -15,7 +15,8 @@ def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument(
         "--dataset_dir",
-        default=(project_path / "data/llm-log-hk/extract-dataset").as_posix(),
+        # default=(project_path / "data/llm-log-hk/extract-dataset").as_posix(),
+        default=(project_path / "data/llm-log-hk/extract-dataset/20250804").as_posix(),
         # default=(project_path / "data/llm-log-usa/extract-dataset").as_posix(),
         type=str
     )
@@ -33,7 +34,11 @@ def main():
     json_dir = dataset_dir / "json"
     json_dir.mkdir(parents=True, exist_ok=True)
 
-    for data_dir in [raw_dir, choice_dir, json_dir]:
+    for data_dir in [
+        raw_dir,
+        choice_dir,
+        json_dir,
+    ]:
         for sample_dir in data_dir.glob("*"):
             # print(f"process: {sample_dir.as_posix()}")
 
@@ -49,10 +54,15 @@ def main():
                 response = f.read()
 
             for key_str in [
+                # "LingoAce",
                 # "BingoPlus",
-                " COD ",
+                # " COD ",
+                # "机器狗",
+                # "andy",
+                "HomePro",
                 # "NXPay",
                 # "NX Money",
+                # "Exodus Bank",
                 # "Exodus Telecom",
                 # "Exodus Retail",
                 # "Exodus Automotive",
@@ -60,19 +70,27 @@ def main():
                 # "NXCloud",
                 # "作为VIP客户",
                 # "FedEx",
+                # "Chinese laser cutting",
             ]:
                 if system_prompt.__contains__(key_str) or user_prompt.__contains__(key_str):
                     print(f"process: {sample_dir.as_posix()}")
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-lingoace"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-bingoplus"
-                    tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-cod"
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-cod"
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-andy"
+                    tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-homepro"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-nxpay"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-nxmoney"
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-exodus-bank"
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-exodus-telecom"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-exodus-retail"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-exodus-automotive"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-kta"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-nxcloud"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-vip"
                     # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-fedex"
+                    # tgt_dir = dataset_dir / f"{data_dir.parts[-1]}-laser"
+
                     tgt_dir.mkdir(parents=True, exist_ok=True)
                     shutil.move(
                         sample_dir.as_posix(),

examples/make_raw_dataset/step_6_filter_by_choice.py

@@ -12,7 +12,7 @@ def get_args():
     parser = argparse.ArgumentParser()
     parser.add_argument(
         "--data_dir",
-        default=(project_path / "data/llm-log-hk/extract-dataset/choice-bingoplus-filipino").as_posix(),
+        default=(project_path / "data/llm-log-hk/extract-dataset/choice-nxpay").as_posix(),
         type=str
     )
     args = parser.parse_args()

examples/test_metrics/bingoplus_chat_metric.py

@@ -38,12 +38,12 @@ python3 azure_openai.py --model_name gpt-4o-mini \
     )
     parser.add_argument(
         "--eval_data_file",
-        default=(project_path / "data/eval_data/gemini_google/google/llama-4-scout-17b-16e-instruct-maas/shenzhen_sase/google_potent_veld_462405_t3/20250731_162116/agent-bingoplus-ph-200-chat.jsonl.raw").as_posix(),
+        default=(project_path / "data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-200-chat.jsonl.raw").as_posix(),
         type=str
     )
     parser.add_argument(
         "--output_file",
-        default=(project_path / "data/eval_data/gemini_google/google/llama-4-scout-17b-16e-instruct-maas/shenzhen_sase/google_potent_veld_462405_t3/20250731_162116/agent-bingoplus-ph-200-chat.jsonl").as_posix(),
+        default=(project_path / "data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-bingoplus-ph-200-chat.jsonl").as_posix(),
         type=str
     )
     parser.add_argument(

examples/test_metrics/lingoace_chat_metric.py

@@ -43,12 +43,12 @@ python3 azure_openai.py --model_name gpt-4o-mini \
     )
     parser.add_argument(
         "--eval_data_file",
-        default=(project_path / "data/eval_data/gemini_google/google/llama-4-scout-17b-16e-instruct-maas/shenzhen_sase/google_potent_veld_462405_t3/20250731_162116/agent-lingoace-zh-80-chat.jsonl.raw").as_posix(),
+        default=(project_path / "data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-80-chat.jsonl.raw").as_posix(),
         type=str
     )
     parser.add_argument(
         "--output_file",
-        default=(project_path / "data/eval_data/gemini_google/google/llama-4-scout-17b-16e-instruct-maas/shenzhen_sase/google_potent_veld_462405_t3/20250731_162116/agent-lingoace-zh-80-chat.jsonl").as_posix(),
+        default=(project_path / "data/eval_data/siliconflow/siliconflow/Tongyi-Zhiwen#QwenLong-L1-32B/shenzhen_sase/siliconflow_api_key/20250814_134104/agent-lingoace-zh-80-chat.jsonl").as_posix(),
         type=str
     )
     parser.add_argument(

llm_eval_script/siliconflow.py

@@ -70,12 +70,19 @@ def get_args():
         "--model_name",
         # default="Pro/deepseek-ai/DeepSeek-R1",
         # default="tencent/Hunyuan-A13B-Instruct",
-        default="deepseek-ai/DeepSeek-V3",
-        # default="Qwen/Qwen3-8B",
+        # default="deepseek-ai/DeepSeek-V3",
         # default="deepseek-ai/DeepSeek-R1",
         # default="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
         # default="deepseek-ai/DeepSeek-R1-Distill-Qwen-7B",
         # default="baidu/ERNIE-4.5-300B-A47B",
+        # default="Qwen/Qwen3-8B",
+        # default="Qwen/Qwen3-14B",
+        # default="Qwen/Qwen3-32B",
+        # default="Qwen/Qwen3-30B-A3B",
+        # default="Qwen/Qwen3-30B-A3B-Instruct-2507",
+        # default="Qwen/Qwen3-235B-A22B-Instruct-2507",
+        # default="Qwen/QwQ-32B",
+        default="Tongyi-Zhiwen/QwenLong-L1-32B",
         type=str
     )
     parser.add_argument(
@@ -108,7 +115,7 @@ def get_args():
     parser.add_argument(
         "--create_time_str",
         default="null",
-        # default="20250728_113641",
+        # default="20250812_092418",
         type=str
     )
     parser.add_argument(

llm_eval_script/siliconflow_chat.py

@@ -51,6 +51,7 @@ from datetime import datetime
 import json
 import os
 from pathlib import Path
+import re
 import sys
 import time
 from zoneinfo import ZoneInfo  # Python 3.9+ 自带，无需安装
@@ -69,12 +70,19 @@ def get_args():
         "--model_name",
         # default="Pro/deepseek-ai/DeepSeek-R1",
         # default="tencent/Hunyuan-A13B-Instruct",
-        default="deepseek-ai/DeepSeek-V3",
-        # default="Qwen/Qwen3-8B",
+        # default="deepseek-ai/DeepSeek-V3",
         # default="deepseek-ai/DeepSeek-R1",
         # default="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
         # default="deepseek-ai/DeepSeek-R1-Distill-Qwen-7B",
         # default="baidu/ERNIE-4.5-300B-A47B",
+        # default="Qwen/Qwen3-8B",
+        # default="Qwen/Qwen3-14B",
+        # default="Qwen/Qwen3-32B",
+        # default="Qwen/Qwen3-30B-A3B",
+        # default="Qwen/Qwen3-30B-A3B-Instruct-2507",
+        # default="Qwen/Qwen3-235B-A22B-Instruct-2507",
+        # default="Qwen/QwQ-32B",
+        default="Tongyi-Zhiwen/QwenLong-L1-32B",
         type=str
     )
     parser.add_argument(
@@ -105,13 +113,13 @@ def get_args():
     )
     parser.add_argument(
         "--create_time_str",
-        default="null",
-        # default="20250728_113641",
+        # default="null",
+        default="20250814_134104",
         type=str
     )
     parser.add_argument(
         "--interval",
-        default=1,
+        default=10,
         type=int
     )
     args = parser.parse_args()
@@ -171,15 +179,43 @@ def main():
                 continue
             finished_idx_set.add(idx)
 
+            # prompt
+            splits = prompt[::-1].split("\n\n", maxsplit=1)
+            conversation = splits[0]
+            system_prompt = splits[1]
+            conversation = conversation[::-1].strip()
+            system_prompt = system_prompt[::-1].strip()
+
+            pattern = "^(Client|Assistant): (.*?)(?=\n(?:Client|Assistant):)"
+            match = re.findall(pattern=pattern, string=conversation, flags=re.I|re.DOTALL|re.MULTILINE)
+
+            messages_ = list()
+            for m in match:
+                role = m[0].lower()
+                content = m[1]
+                if role in ("client", "Client"):
+                    role = "user"
+                elif role in ("assistant", "Assistant"):
+                    role = "assistant"
+                else:
+                    raise AssertionError
+                messages_.append({
+                    "role": role,
+                    "content": content
+                })
+
+            messages = [
+                {"role": "system", "content": system_prompt},
+                *messages_
+            ]
+
             try:
                 time.sleep(args.interval)
                 print(f"sleep: {args.interval}")
                 time_begin = time.time()
                 completion = client.chat.completions.create(
                     model=args.model_name,
-                    messages=[
-                        {"role": "user", "content": prompt},
-                    ],
+                    messages=messages,
                     stream=False,
                     max_tokens=4096,
                     # max_tokens=1,

llm_eval_script/siliconflow_summary.py

@@ -69,12 +69,19 @@ def get_args():
         "--model_name",
         # default="Pro/deepseek-ai/DeepSeek-R1",
         # default="tencent/Hunyuan-A13B-Instruct",
-        default="deepseek-ai/DeepSeek-V3",
-        # default="Qwen/Qwen3-8B",
+        # default="deepseek-ai/DeepSeek-V3",
         # default="deepseek-ai/DeepSeek-R1",
         # default="deepseek-ai/DeepSeek-R1-0528-Qwen3-8B",
         # default="deepseek-ai/DeepSeek-R1-Distill-Qwen-7B",
         # default="baidu/ERNIE-4.5-300B-A47B",
+        # default="Qwen/Qwen3-8B",
+        # default="Qwen/Qwen3-14B",
+        default="Qwen/Qwen3-32B",
+        # default="Qwen/Qwen3-30B-A3B",
+        # default="Qwen/Qwen3-30B-A3B-Instruct-2507",
+        # default="Qwen/Qwen3-235B-A22B-Instruct-2507",
+        # default="Qwen/QwQ-32B",
+        # default="Tongyi-Zhiwen/QwenLong-L1-32B",
         type=str
     )
     parser.add_argument(