deeplumiere commited on
Commit
ae44f4c
·
verified ·
1 Parent(s): e0e0021

Add files using upload-large-folder tool

Browse files
This view is limited to 50 files because it contains too many changes.   See raw diff
Files changed (50) hide show
  1. data/hal/hal-corebench-hard/google/gemini-2-0-flash/27a6cc1d-b40f-4bd1-b3fa-9b7757b3bb21.json +90 -0
  2. data/hal/hal-corebench-hard/google/gemini-2-0-flash/e7784822-6bc6-405f-837d-3771efe67e9f.json +90 -0
  3. data/hal/hal-corebench-hard/google/gemini-2-5-pro-preview/853a7e34-106e-4a97-8098-23ced4724ebb.json +90 -0
  4. data/hal/hal-corebench-hard/google/gemini-2-5-pro-preview/ded7ca38-d2dc-4965-b68d-ca914e3407bf.json +90 -0
  5. data/hal/hal-corebench-hard/google/gemini-3-pro-preview/9d4f2418-11f0-496f-ad16-7ce24f0074fc.json +91 -0
  6. data/hal/hal-corebench-hard/google/gemini-3-pro-preview/c155b4ea-fab4-4fc3-b9c0-2f4cb94b67b8.json +91 -0
  7. data/hal/hal-corebench-hard/openai/o3/0da61ef1-7196-4c35-85e7-fded635e2712.json +91 -0
  8. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/08702d9e-cf6e-41ad-ab9f-c63697151201.json +82 -0
  9. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/1b6d5141-35e5-4cb7-b7b2-0177e4df3334.json +83 -0
  10. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/3cf52138-2ed6-47cb-b695-f8e4d2dd2029.json +82 -0
  11. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/7ab3a7f2-11e4-4117-99e4-d353f8530c19.json +82 -0
  12. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/7d9b6419-f288-4c7f-95b4-c5616e64835c.json +83 -0
  13. data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/87584480-5776-4227-a4c5-b08b5c59758a.json +83 -0
  14. data/hal/hal-scicode/anthropic/claude-haiku-4-5/a8949ff9-d2cc-4fa8-9b13-77c8584a5ae1.json +82 -0
  15. data/hal/hal-scicode/anthropic/claude-opus-4-1/2c492364-474e-4212-ae8b-ebe9ce8cfe9f.json +85 -0
  16. data/hal/hal-scicode/anthropic/claude-opus-4-1/c3233b51-d181-46a3-8bda-86c5a76bec48.json +82 -0
  17. data/hal/hal-scicode/anthropic/claude-sonnet-4-5/3441cdb6-9242-41ae-8dee-f6ab5aa93d6a.json +82 -0
  18. data/hal/hal-scicode/anthropic/claude-sonnet-4-5/7231dfdd-9ca5-4e72-abcf-15b1f6b0337a.json +83 -0
  19. data/hal/hal-scicode/deepseek/deepseek-r1/48a88cf2-9f82-46c1-b8c6-cc8c87a54b7f.json +82 -0
  20. data/hal/hal-scicode/deepseek/deepseek-r1/51ff8963-b93c-458c-9e0b-729076da3adc.json +82 -0
  21. data/hal/hal-scicode/deepseek/deepseek-r1/9f4899c6-1eaf-4153-b33c-1950deb5111c.json +82 -0
  22. data/hal/hal-scicode/deepseek/deepseek-v3/3704f223-ea1b-49b3-ba09-8a2eb68a25ca.json +82 -0
  23. data/hal/hal-scicode/deepseek/deepseek-v3/4170847c-e2e7-49b2-af4d-ddea0847ab9e.json +82 -0
  24. data/hal/hal-scicode/deepseek/deepseek-v3/7cad859f-ba86-4b55-ad66-3d423646a3a6.json +82 -0
  25. data/hal/hal-scicode/google/gemini-2-0-flash/73f62d6e-9b7d-4c0a-8bb8-083d143fc405.json +82 -0
  26. data/hal/hal-scicode/google/gemini-2-0-flash/9c3c4d2a-7ceb-43ba-b630-f5ece60aacff.json +82 -0
  27. data/hal/hal-scicode/google/gemini-2-0-flash/eb742aa4-2612-4e43-a819-ed09339ac4bc.json +82 -0
  28. data/hal/hal-scicode/openai/gpt-4-1/0a4b4d3b-735d-482a-8a69-e5922c7e5ce2.json +82 -0
  29. data/hal/hal-scicode/openai/gpt-4-1/5ec554d6-c0b1-4d22-8c9a-82291604239f.json +82 -0
  30. data/hal/hal-scicode/openai/gpt-4-1/f97a3701-0178-4c71-a10b-a9932e4a9c5c.json +82 -0
  31. data/hal/hal-scicode/openai/gpt-5/e800b2f1-841e-4f1e-8d0c-de7af622b09a.json +83 -0
  32. data/hal/hal-scicode/openai/o3/14940413-7ee1-46d3-bf6f-4090afa08aa9.json +83 -0
  33. data/hal/hal-scicode/openai/o3/22b79ac8-600c-48f7-bcaf-711bb853d247.json +83 -0
  34. data/hal/hal-scicode/openai/o3/fbe6ea98-e3da-4049-b08d-42088f2781e6.json +83 -0
  35. data/hal/hal-scicode/openai/o4-mini/0be4c080-4af7-4e0d-80db-1b161f732e76.json +83 -0
  36. data/hal/hal-scicode/openai/o4-mini/27e60cf9-5174-47bd-aaf1-2604664186ec.json +83 -0
  37. data/hal/hal-scicode/openai/o4-mini/43967556-fc28-41fa-bf1c-bed6510b638e.json +83 -0
  38. data/hal/hal-scicode/openai/o4-mini/4716f4a5-5cb9-4a3f-aa2d-530e4a2e4efe.json +83 -0
  39. data/hal/hal-scicode/openai/o4-mini/efd7e370-8418-4055-acfa-dbe916785995.json +83 -0
  40. data/hal/hal-scicode/openai/o4-mini/f064dd1c-e6b5-4a8a-8b60-afe978cf929e.json +83 -0
  41. data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/05495528-2651-424c-923d-32e200921b73.json +87 -0
  42. data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/060145ef-ace8-47a2-a59a-f702d87c4d98.json +86 -0
  43. data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/10384a3b-43e9-4ace-a96e-a823852a8426.json +86 -0
  44. data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/391024b2-83e3-49eb-a8ef-6161cdfe7405.json +87 -0
  45. data/hal/hal-swebench-verified-mini/anthropic/claude-haiku-4-5/7c6a02a4-1d18-4368-9451-c770584abead.json +87 -0
  46. data/hal/hal-swebench-verified-mini/anthropic/claude-haiku-4-5/b417daf8-88bf-4c53-837d-9745b3d8dc8d.json +88 -0
  47. data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/08d3af13-82c1-4710-bae1-a7b1837189ca.json +88 -0
  48. data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/a74f18a1-bc74-491e-b9ac-a365e5f22dcc.json +86 -0
  49. data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/dd76a080-eb8e-4161-b709-99132ea09733.json +87 -0
  50. data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/f2746072-cb8f-4efd-9b1a-3dc74be1ee38.json +87 -0
data/hal/hal-corebench-hard/google/gemini-2-0-flash/27a6cc1d-b40f-4bd1-b3fa-9b7757b3bb21.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/hal-generalist-agent/gemini-2-0-flash-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.0 Flash (February 2025)",
23
+ "id": "google/gemini-2.0-flash",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.0 Flash (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "total_cost_usd": "7.06"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "CORE-Bench Hard",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "CORE-Bench Hard",
38
+ "url": [
39
+ "https://github.com/siegelz/core-bench",
40
+ "https://hal.cs.princeton.edu/corebench_hard"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0444,
52
+ "details": {
53
+ "accuracy_raw": "4.44%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "python",
66
+ "description": "Execute Python code"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the filesystem"
71
+ },
72
+ {
73
+ "name": "write_file",
74
+ "description": "Write files to the filesystem"
75
+ }
76
+ ]
77
+ }
78
+ },
79
+ "additional_details": {
80
+ "agent_scaffold": "HAL Generalist Agent",
81
+ "hal_rank": "48",
82
+ "runs": "1",
83
+ "verified": "True",
84
+ "is_pareto": "False",
85
+ "total_cost_usd": "7.06"
86
+ }
87
+ }
88
+ }
89
+ ]
90
+ }
data/hal/hal-corebench-hard/google/gemini-2-0-flash/e7784822-6bc6-405f-837d-3771efe67e9f.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/core-agent/gemini-2-0-flash-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.0 Flash (February 2025)",
23
+ "id": "google/gemini-2.0-flash",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.0 Flash (February 2025)",
27
+ "agent_scaffold": "CORE-Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "total_cost_usd": "12.46"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "CORE-Bench Hard",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "CORE-Bench Hard",
38
+ "url": [
39
+ "https://github.com/siegelz/core-bench",
40
+ "https://hal.cs.princeton.edu/corebench_hard"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.1111,
52
+ "details": {
53
+ "accuracy_raw": "11.11%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "python",
66
+ "description": "Execute Python code"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the filesystem"
71
+ },
72
+ {
73
+ "name": "write_file",
74
+ "description": "Write files to the filesystem"
75
+ }
76
+ ]
77
+ }
78
+ },
79
+ "additional_details": {
80
+ "agent_scaffold": "CORE-Agent",
81
+ "hal_rank": "39",
82
+ "runs": "1",
83
+ "verified": "True",
84
+ "is_pareto": "False",
85
+ "total_cost_usd": "12.46"
86
+ }
87
+ }
88
+ }
89
+ ]
90
+ }
data/hal/hal-corebench-hard/google/gemini-2-5-pro-preview/853a7e34-106e-4a97-8098-23ced4724ebb.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/hal-generalist-agent/gemini-2-5-pro-preview-march-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.5 Pro Preview (March 2025)",
23
+ "id": "google/gemini-2.5-pro-preview",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.5 Pro Preview (March 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "total_cost_usd": "30.38"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "CORE-Bench Hard",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "CORE-Bench Hard",
38
+ "url": [
39
+ "https://github.com/siegelz/core-bench",
40
+ "https://hal.cs.princeton.edu/corebench_hard"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0444,
52
+ "details": {
53
+ "accuracy_raw": "4.44%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "python",
66
+ "description": "Execute Python code"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the filesystem"
71
+ },
72
+ {
73
+ "name": "write_file",
74
+ "description": "Write files to the filesystem"
75
+ }
76
+ ]
77
+ }
78
+ },
79
+ "additional_details": {
80
+ "agent_scaffold": "HAL Generalist Agent",
81
+ "hal_rank": "49",
82
+ "runs": "1",
83
+ "verified": "True",
84
+ "is_pareto": "False",
85
+ "total_cost_usd": "30.38"
86
+ }
87
+ }
88
+ }
89
+ ]
90
+ }
data/hal/hal-corebench-hard/google/gemini-2-5-pro-preview/ded7ca38-d2dc-4965-b68d-ca914e3407bf.json ADDED
@@ -0,0 +1,90 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/core-agent/gemini-2-5-pro-preview-march-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.5 Pro Preview (March 2025)",
23
+ "id": "google/gemini-2.5-pro-preview",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.5 Pro Preview (March 2025)",
27
+ "agent_scaffold": "CORE-Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "total_cost_usd": "182.34"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "CORE-Bench Hard",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "CORE-Bench Hard",
38
+ "url": [
39
+ "https://github.com/siegelz/core-bench",
40
+ "https://hal.cs.princeton.edu/corebench_hard"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.2222,
52
+ "details": {
53
+ "accuracy_raw": "22.22%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "python",
66
+ "description": "Execute Python code"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the filesystem"
71
+ },
72
+ {
73
+ "name": "write_file",
74
+ "description": "Write files to the filesystem"
75
+ }
76
+ ]
77
+ }
78
+ },
79
+ "additional_details": {
80
+ "agent_scaffold": "CORE-Agent",
81
+ "hal_rank": "32",
82
+ "runs": "1",
83
+ "verified": "True",
84
+ "is_pareto": "False",
85
+ "total_cost_usd": "182.34"
86
+ }
87
+ }
88
+ }
89
+ ]
90
+ }
data/hal/hal-corebench-hard/google/gemini-3-pro-preview/9d4f2418-11f0-496f-ad16-7ce24f0074fc.json ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/hal-generalist-agent/gemini-3-pro-preview-high-november-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 3 Pro Preview High (November 2025)",
23
+ "id": "google/gemini-3-pro-preview",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 3 Pro Preview High (November 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "101.27"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "CORE-Bench Hard",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "CORE-Bench Hard",
39
+ "url": [
40
+ "https://github.com/siegelz/core-bench",
41
+ "https://hal.cs.princeton.edu/corebench_hard"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.3556,
53
+ "details": {
54
+ "accuracy_raw": "35.56%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "python",
67
+ "description": "Execute Python code"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the filesystem"
72
+ },
73
+ {
74
+ "name": "write_file",
75
+ "description": "Write files to the filesystem"
76
+ }
77
+ ]
78
+ }
79
+ },
80
+ "additional_details": {
81
+ "agent_scaffold": "HAL Generalist Agent",
82
+ "hal_rank": "15",
83
+ "runs": "1",
84
+ "verified": "True",
85
+ "is_pareto": "False",
86
+ "total_cost_usd": "101.27"
87
+ }
88
+ }
89
+ }
90
+ ]
91
+ }
data/hal/hal-corebench-hard/google/gemini-3-pro-preview/c155b4ea-fab4-4fc3-b9c0-2f4cb94b67b8.json ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/core-agent/gemini-3-pro-preview-high-november-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 3 Pro Preview High (November 2025)",
23
+ "id": "google/gemini-3-pro-preview",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 3 Pro Preview High (November 2025)",
27
+ "agent_scaffold": "CORE-Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "86.6"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "CORE-Bench Hard",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "CORE-Bench Hard",
39
+ "url": [
40
+ "https://github.com/siegelz/core-bench",
41
+ "https://hal.cs.princeton.edu/corebench_hard"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.4,
53
+ "details": {
54
+ "accuracy_raw": "40.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "python",
67
+ "description": "Execute Python code"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the filesystem"
72
+ },
73
+ {
74
+ "name": "write_file",
75
+ "description": "Write files to the filesystem"
76
+ }
77
+ ]
78
+ }
79
+ },
80
+ "additional_details": {
81
+ "agent_scaffold": "CORE-Agent",
82
+ "hal_rank": "10",
83
+ "runs": "1",
84
+ "verified": "True",
85
+ "is_pareto": "False",
86
+ "total_cost_usd": "86.6"
87
+ }
88
+ }
89
+ }
90
+ ]
91
+ }
data/hal/hal-corebench-hard/openai/o3/0da61ef1-7196-4c35-85e7-fded635e2712.json ADDED
@@ -0,0 +1,91 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-corebench-hard/hal-generalist-agent/o3-medium-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — CORE-Bench Hard",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "corebench_hard"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o3 Medium (April 2025)",
23
+ "id": "openai/o3",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o3 Medium (April 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "CORE-Bench Hard",
29
+ "inference_effort": "medium",
30
+ "total_cost_usd": "88.34"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "CORE-Bench Hard",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "CORE-Bench Hard",
39
+ "url": [
40
+ "https://github.com/siegelz/core-bench",
41
+ "https://hal.cs.princeton.edu/corebench_hard"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of CORE-Bench Hard tasks solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.2222,
53
+ "details": {
54
+ "accuracy_raw": "22.22%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "python",
67
+ "description": "Execute Python code"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the filesystem"
72
+ },
73
+ {
74
+ "name": "write_file",
75
+ "description": "Write files to the filesystem"
76
+ }
77
+ ]
78
+ }
79
+ },
80
+ "additional_details": {
81
+ "agent_scaffold": "HAL Generalist Agent",
82
+ "hal_rank": "31",
83
+ "runs": "1",
84
+ "verified": "True",
85
+ "is_pareto": "False",
86
+ "total_cost_usd": "88.34"
87
+ }
88
+ }
89
+ }
90
+ ]
91
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/08702d9e-cf6e-41ad-ab9f-c63697151201.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/claude-3-7-sonnet-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "60.4"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0308,
52
+ "details": {
53
+ "accuracy_raw": "3.08%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "HAL Generalist Agent",
73
+ "hal_rank": "15",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "60.4"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/1b6d5141-35e5-4cb7-b7b2-0177e4df3334.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/claude-3-7-sonnet-high-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet High (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet High (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "188.15"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0308,
53
+ "details": {
54
+ "accuracy_raw": "3.08%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "HAL Generalist Agent",
74
+ "hal_rank": "18",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "188.15"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/3cf52138-2ed6-47cb-b695-f8e4d2dd2029.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-3-7-sonnet-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet (February 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "191.41"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0308,
52
+ "details": {
53
+ "accuracy_raw": "3.08%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "19",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "191.41"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/7ab3a7f2-11e4-4117-99e4-d353f8530c19.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/claude-3-7-sonnet-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet (February 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "5.1"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Zero Shot Agent",
73
+ "hal_rank": "27",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "5.1"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/7d9b6419-f288-4c7f-95b4-c5616e64835c.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-3-7-sonnet-high-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet High (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet High (February 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "204.37"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0462,
53
+ "details": {
54
+ "accuracy_raw": "4.62%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "12",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "204.37"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/anthropic/claude-3-7-sonnet-20250219/87584480-5776-4227-a4c5-b08b5c59758a.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/claude-3-7-sonnet-high-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet High (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet High (February 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "4.99"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0308,
53
+ "details": {
54
+ "accuracy_raw": "3.08%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Zero Shot Agent",
74
+ "hal_rank": "14",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "4.99"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/anthropic/claude-haiku-4-5/a8949ff9-d2cc-4fa8-9b13-77c8584a5ae1.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-haiku-4-5-october-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Haiku 4.5 (October 2025)",
23
+ "id": "anthropic/claude-haiku-4.5",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Haiku 4.5 (October 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "232.36"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "32",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "232.36"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-opus-4-1/2c492364-474e-4212-ae8b-ebe9ce8cfe9f.json ADDED
@@ -0,0 +1,85 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-opus-4-1-high-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 High (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 High (August 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "550.54"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0692,
53
+ "details": {
54
+ "accuracy_raw": "6.92%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "4",
75
+ "runs": "2",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "550.54",
79
+ "cost_confidence_interval": "-42.81/+42.81",
80
+ "accuracy_confidence_interval": "-0.77/+0.77"
81
+ }
82
+ }
83
+ }
84
+ ]
85
+ }
data/hal/hal-scicode/anthropic/claude-opus-4-1/c3233b51-d181-46a3-8bda-86c5a76bec48.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-opus-4-1-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 (August 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "625.13"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0769,
52
+ "details": {
53
+ "accuracy_raw": "7.69%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "3",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "625.13"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-sonnet-4-5/3441cdb6-9242-41ae-8dee-f6ab5aa93d6a.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-sonnet-4-5-september-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Sonnet 4.5 (September 2025)",
23
+ "id": "anthropic/claude-sonnet-4-5",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Sonnet 4.5 (September 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "104.29"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0308,
52
+ "details": {
53
+ "accuracy_raw": "3.08%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "17",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "104.29"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/anthropic/claude-sonnet-4-5/7231dfdd-9ca5-4e72-abcf-15b1f6b0337a.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/claude-sonnet-4-5-high-september-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Sonnet 4.5 High (September 2025)",
23
+ "id": "anthropic/claude-sonnet-4-5",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Sonnet 4.5 High (September 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "118.14"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0154,
53
+ "details": {
54
+ "accuracy_raw": "1.54%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "25",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "118.14"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/deepseek/deepseek-r1/48a88cf2-9f82-46c1-b8c6-cc8c87a54b7f.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/deepseek-r1-january-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek R1 (January 2025)",
23
+ "id": "deepseek/deepseek-r1",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek R1 (January 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "486.78"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "HAL Generalist Agent",
73
+ "hal_rank": "33",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "486.78"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/deepseek/deepseek-r1/51ff8963-b93c-458c-9e0b-729076da3adc.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/deepseek-r1-may-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek R1 (May 2025)",
23
+ "id": "deepseek/deepseek-r1",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek R1 (May 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "2.19"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Zero Shot Agent",
73
+ "hal_rank": "26",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "2.19"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/deepseek/deepseek-r1/9f4899c6-1eaf-4153-b33c-1950deb5111c.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/deepseek-r1-may-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek R1 (May 2025)",
23
+ "id": "deepseek/deepseek-r1",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek R1 (May 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "57.62"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "29",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "57.62"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/deepseek/deepseek-v3/3704f223-ea1b-49b3-ba09-8a2eb68a25ca.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/deepseek-v3-march-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek V3 (March 2025)",
23
+ "id": "deepseek/deepseek-v3",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek V3 (March 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "52.11"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "28",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "52.11"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/deepseek/deepseek-v3/4170847c-e2e7-49b2-af4d-ddea0847ab9e.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/deepseek-v3-march-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek V3 (March 2025)",
23
+ "id": "deepseek/deepseek-v3",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek V3 (March 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "219.36"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "HAL Generalist Agent",
73
+ "hal_rank": "31",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "219.36"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/deepseek/deepseek-v3/7cad859f-ba86-4b55-ad66-3d423646a3a6.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/deepseek-v3-march-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "DeepSeek V3 (March 2025)",
23
+ "id": "deepseek/deepseek-v3",
24
+ "developer": "deepseek",
25
+ "additional_details": {
26
+ "hal_model_name": "DeepSeek V3 (March 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "0.79"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0308,
52
+ "details": {
53
+ "accuracy_raw": "3.08%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Zero Shot Agent",
73
+ "hal_rank": "13",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "0.79"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/google/gemini-2-0-flash/73f62d6e-9b7d-4c0a-8bb8-083d143fc405.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/gemini-2-0-flash-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.0 Flash (February 2025)",
23
+ "id": "google/gemini-2.0-flash",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.0 Flash (February 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "5.23"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0154,
52
+ "details": {
53
+ "accuracy_raw": "1.54%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "21",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "5.23"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/google/gemini-2-0-flash/9c3c4d2a-7ceb-43ba-b630-f5ece60aacff.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/gemini-2-0-flash-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.0 Flash (February 2025)",
23
+ "id": "google/gemini-2.0-flash",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.0 Flash (February 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "0.12"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0154,
52
+ "details": {
53
+ "accuracy_raw": "1.54%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Zero Shot Agent",
73
+ "hal_rank": "20",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "True",
77
+ "total_cost_usd": "0.12"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/google/gemini-2-0-flash/eb742aa4-2612-4e43-a819-ed09339ac4bc.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/gemini-2-0-flash-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Gemini 2.0 Flash (February 2025)",
23
+ "id": "google/gemini-2.0-flash",
24
+ "developer": "google",
25
+ "additional_details": {
26
+ "hal_model_name": "Gemini 2.0 Flash (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "61.49"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0,
52
+ "details": {
53
+ "accuracy_raw": "0.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "HAL Generalist Agent",
73
+ "hal_rank": "30",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "61.49"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/openai/gpt-4-1/0a4b4d3b-735d-482a-8a69-e5922c7e5ce2.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/gpt-4-1-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "GPT-4.1 (April 2025)",
23
+ "id": "openai/gpt-4.1",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "GPT-4.1 (April 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "69.39"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0154,
52
+ "details": {
53
+ "accuracy_raw": "1.54%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Tool Calling Agent",
73
+ "hal_rank": "22",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "69.39"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/openai/gpt-4-1/5ec554d6-c0b1-4d22-8c9a-82291604239f.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/gpt-4-1-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "GPT-4.1 (April 2025)",
23
+ "id": "openai/gpt-4.1",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "GPT-4.1 (April 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "73.87"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0154,
52
+ "details": {
53
+ "accuracy_raw": "1.54%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "HAL Generalist Agent",
73
+ "hal_rank": "23",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "73.87"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/openai/gpt-4-1/f97a3701-0178-4c71-a10b-a9932e4a9c5c.json ADDED
@@ -0,0 +1,82 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/gpt-4-1-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "GPT-4.1 (April 2025)",
23
+ "id": "openai/gpt-4.1",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "GPT-4.1 (April 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "total_cost_usd": "2.82"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "Scicode",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "Scicode",
38
+ "url": [
39
+ "https://scicode-bench.github.io",
40
+ "https://hal.cs.princeton.edu/scicode"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.0615,
52
+ "details": {
53
+ "accuracy_raw": "6.15%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "python",
62
+ "description": "Execute Python code for scientific computation"
63
+ },
64
+ {
65
+ "name": "bash",
66
+ "description": "Execute shell commands"
67
+ }
68
+ ]
69
+ }
70
+ },
71
+ "additional_details": {
72
+ "agent_scaffold": "Scicode Zero Shot Agent",
73
+ "hal_rank": "5",
74
+ "runs": "1",
75
+ "verified": "True",
76
+ "is_pareto": "False",
77
+ "total_cost_usd": "2.82"
78
+ }
79
+ }
80
+ }
81
+ ]
82
+ }
data/hal/hal-scicode/openai/gpt-5/e800b2f1-841e-4f1e-8d0c-de7af622b09a.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/gpt-5-medium-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "GPT-5 Medium (August 2025)",
23
+ "id": "openai/gpt-5",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "GPT-5 Medium (August 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "medium",
30
+ "total_cost_usd": "193.52"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0615,
53
+ "details": {
54
+ "accuracy_raw": "6.15%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "8",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "193.52"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o3/14940413-7ee1-46d3-bf6f-4090afa08aa9.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/o3-medium-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o3 Medium (April 2025)",
23
+ "id": "openai/o3",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o3 Medium (April 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "medium",
30
+ "total_cost_usd": "66.98"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0308,
53
+ "details": {
54
+ "accuracy_raw": "3.08%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "HAL Generalist Agent",
74
+ "hal_rank": "16",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "66.98"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o3/22b79ac8-600c-48f7-bcaf-711bb853d247.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/o3-medium-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o3 Medium (April 2025)",
23
+ "id": "openai/o3",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o3 Medium (April 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "medium",
30
+ "total_cost_usd": "6.03"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0462,
53
+ "details": {
54
+ "accuracy_raw": "4.62%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Zero Shot Agent",
74
+ "hal_rank": "9",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "6.03"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o3/fbe6ea98-e3da-4049-b08d-42088f2781e6.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/o3-medium-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o3 Medium (April 2025)",
23
+ "id": "openai/o3",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o3 Medium (April 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "medium",
30
+ "total_cost_usd": "111.11"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0923,
53
+ "details": {
54
+ "accuracy_raw": "9.23%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "2",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "111.11"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/0be4c080-4af7-4e0d-80db-1b161f732e76.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/o4-mini-high-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini High (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini High (April 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "66.2"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0462,
53
+ "details": {
54
+ "accuracy_raw": "4.62%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "11",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "66.2"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/27e60cf9-5174-47bd-aaf1-2604664186ec.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/o4-mini-high-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini High (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini High (April 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "5.37"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0615,
53
+ "details": {
54
+ "accuracy_raw": "6.15%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Zero Shot Agent",
74
+ "hal_rank": "6",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "5.37"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/43967556-fc28-41fa-bf1c-bed6510b638e.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-tool-calling-agent/o4-mini-low-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini Low (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini Low (April 2025)",
27
+ "agent_scaffold": "Scicode Tool Calling Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "low",
30
+ "total_cost_usd": "46.3"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0462,
53
+ "details": {
54
+ "accuracy_raw": "4.62%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Tool Calling Agent",
74
+ "hal_rank": "10",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "46.3"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/4716f4a5-5cb9-4a3f-aa2d-530e4a2e4efe.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/o4-mini-high-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini High (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini High (April 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "92.1"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0154,
53
+ "details": {
54
+ "accuracy_raw": "1.54%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "HAL Generalist Agent",
74
+ "hal_rank": "24",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "92.1"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/efd7e370-8418-4055-acfa-dbe916785995.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/scicode-zero-shot-agent/o4-mini-low-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini Low (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini Low (April 2025)",
27
+ "agent_scaffold": "Scicode Zero Shot Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "low",
30
+ "total_cost_usd": "1.74"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0923,
53
+ "details": {
54
+ "accuracy_raw": "9.23%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "Scicode Zero Shot Agent",
74
+ "hal_rank": "1",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "True",
78
+ "total_cost_usd": "1.74"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-scicode/openai/o4-mini/f064dd1c-e6b5-4a8a-8b60-afe978cf929e.json ADDED
@@ -0,0 +1,83 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-scicode/hal-generalist-agent/o4-mini-low-april-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — Scicode",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Scientific Programming",
14
+ "benchmark_slug": "scicode"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "o4-mini Low (April 2025)",
23
+ "id": "openai/o4-mini",
24
+ "developer": "openai",
25
+ "additional_details": {
26
+ "hal_model_name": "o4-mini Low (April 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "Scicode",
29
+ "inference_effort": "low",
30
+ "total_cost_usd": "165.9"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "Scicode",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "Scicode",
39
+ "url": [
40
+ "https://scicode-bench.github.io",
41
+ "https://hal.cs.princeton.edu/scicode"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of Scicode problems solved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.0615,
53
+ "details": {
54
+ "accuracy_raw": "6.15%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "python",
63
+ "description": "Execute Python code for scientific computation"
64
+ },
65
+ {
66
+ "name": "bash",
67
+ "description": "Execute shell commands"
68
+ }
69
+ ]
70
+ }
71
+ },
72
+ "additional_details": {
73
+ "agent_scaffold": "HAL Generalist Agent",
74
+ "hal_rank": "7",
75
+ "runs": "1",
76
+ "verified": "True",
77
+ "is_pareto": "False",
78
+ "total_cost_usd": "165.9"
79
+ }
80
+ }
81
+ }
82
+ ]
83
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/05495528-2651-424c-923d-32e200921b73.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-3-7-sonnet-high-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet High (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet High (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "72.98"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "SWE-bench Verified Mini",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "SWE-bench Verified Mini",
39
+ "url": [
40
+ "https://www.swebench.com",
41
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.24,
53
+ "details": {
54
+ "accuracy_raw": "24.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "edit_file",
67
+ "description": "Edit files in the repository"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the repository"
72
+ }
73
+ ]
74
+ }
75
+ },
76
+ "additional_details": {
77
+ "agent_scaffold": "HAL Generalist Agent",
78
+ "hal_rank": "23",
79
+ "runs": "1",
80
+ "verified": "True",
81
+ "is_pareto": "False",
82
+ "total_cost_usd": "72.98"
83
+ }
84
+ }
85
+ }
86
+ ]
87
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/060145ef-ace8-47a2-a59a-f702d87c4d98.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/swe-agent/claude-3-7-sonnet-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet (February 2025)",
27
+ "agent_scaffold": "SWE-Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "total_cost_usd": "402.69"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "SWE-bench Verified Mini",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "SWE-bench Verified Mini",
38
+ "url": [
39
+ "https://www.swebench.com",
40
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.5,
52
+ "details": {
53
+ "accuracy_raw": "50.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "edit_file",
66
+ "description": "Edit files in the repository"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the repository"
71
+ }
72
+ ]
73
+ }
74
+ },
75
+ "additional_details": {
76
+ "agent_scaffold": "SWE-Agent",
77
+ "hal_rank": "8",
78
+ "runs": "1",
79
+ "verified": "True",
80
+ "is_pareto": "False",
81
+ "total_cost_usd": "402.69"
82
+ }
83
+ }
84
+ }
85
+ ]
86
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/10384a3b-43e9-4ace-a96e-a823852a8426.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-3-7-sonnet-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet (February 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "total_cost_usd": "117.43"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "SWE-bench Verified Mini",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "SWE-bench Verified Mini",
38
+ "url": [
39
+ "https://www.swebench.com",
40
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.26,
52
+ "details": {
53
+ "accuracy_raw": "26.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "edit_file",
66
+ "description": "Edit files in the repository"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the repository"
71
+ }
72
+ ]
73
+ }
74
+ },
75
+ "additional_details": {
76
+ "agent_scaffold": "HAL Generalist Agent",
77
+ "hal_rank": "20",
78
+ "runs": "1",
79
+ "verified": "True",
80
+ "is_pareto": "False",
81
+ "total_cost_usd": "117.43"
82
+ }
83
+ }
84
+ }
85
+ ]
86
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-3-7-sonnet-20250219/391024b2-83e3-49eb-a8ef-6161cdfe7405.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/swe-agent/claude-3-7-sonnet-high-february-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude-3.7 Sonnet High (February 2025)",
23
+ "id": "anthropic/claude-3-7-sonnet-20250219",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude-3.7 Sonnet High (February 2025)",
27
+ "agent_scaffold": "SWE-Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "388.88"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "SWE-bench Verified Mini",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "SWE-bench Verified Mini",
39
+ "url": [
40
+ "https://www.swebench.com",
41
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.54,
53
+ "details": {
54
+ "accuracy_raw": "54.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "edit_file",
67
+ "description": "Edit files in the repository"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the repository"
72
+ }
73
+ ]
74
+ }
75
+ },
76
+ "additional_details": {
77
+ "agent_scaffold": "SWE-Agent",
78
+ "hal_rank": "5",
79
+ "runs": "1",
80
+ "verified": "True",
81
+ "is_pareto": "False",
82
+ "total_cost_usd": "388.88"
83
+ }
84
+ }
85
+ }
86
+ ]
87
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-haiku-4-5/7c6a02a4-1d18-4368-9451-c770584abead.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-haiku-4-5-high-october-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Haiku 4.5 High (October 2025)",
23
+ "id": "anthropic/claude-haiku-4.5",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Haiku 4.5 High (October 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "65.31"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "SWE-bench Verified Mini",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "SWE-bench Verified Mini",
39
+ "url": [
40
+ "https://www.swebench.com",
41
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.44,
53
+ "details": {
54
+ "accuracy_raw": "44.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "edit_file",
67
+ "description": "Edit files in the repository"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the repository"
72
+ }
73
+ ]
74
+ }
75
+ },
76
+ "additional_details": {
77
+ "agent_scaffold": "HAL Generalist Agent",
78
+ "hal_rank": "13",
79
+ "runs": "1",
80
+ "verified": "True",
81
+ "is_pareto": "True",
82
+ "total_cost_usd": "65.31"
83
+ }
84
+ }
85
+ }
86
+ ]
87
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-haiku-4-5/b417daf8-88bf-4c53-837d-9745b3d8dc8d.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-haiku-4-5-october-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Haiku 4.5 (October 2025)",
23
+ "id": "anthropic/claude-haiku-4.5",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Haiku 4.5 (October 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "total_cost_usd": "147.89"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "SWE-bench Verified Mini",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "SWE-bench Verified Mini",
38
+ "url": [
39
+ "https://www.swebench.com",
40
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.24,
52
+ "details": {
53
+ "accuracy_raw": "24.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "edit_file",
66
+ "description": "Edit files in the repository"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the repository"
71
+ }
72
+ ]
73
+ }
74
+ },
75
+ "additional_details": {
76
+ "agent_scaffold": "HAL Generalist Agent",
77
+ "hal_rank": "24",
78
+ "runs": "2",
79
+ "verified": "True",
80
+ "is_pareto": "False",
81
+ "total_cost_usd": "147.89",
82
+ "cost_confidence_interval": "-60.67/+60.67",
83
+ "accuracy_confidence_interval": "-0.00/+0.00"
84
+ }
85
+ }
86
+ }
87
+ ]
88
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/08d3af13-82c1-4710-bae1-a7b1837189ca.json ADDED
@@ -0,0 +1,88 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/swe-agent/claude-opus-4-1-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 (August 2025)",
27
+ "agent_scaffold": "SWE-Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "total_cost_usd": "1351.35"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "SWE-bench Verified Mini",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "SWE-bench Verified Mini",
38
+ "url": [
39
+ "https://www.swebench.com",
40
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.61,
52
+ "details": {
53
+ "accuracy_raw": "61.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "edit_file",
66
+ "description": "Edit files in the repository"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the repository"
71
+ }
72
+ ]
73
+ }
74
+ },
75
+ "additional_details": {
76
+ "agent_scaffold": "SWE-Agent",
77
+ "hal_rank": "3",
78
+ "runs": "2",
79
+ "verified": "True",
80
+ "is_pareto": "False",
81
+ "total_cost_usd": "1351.35",
82
+ "cost_confidence_interval": "-438.32/+438.32",
83
+ "accuracy_confidence_interval": "-7.00/+7.00"
84
+ }
85
+ }
86
+ }
87
+ ]
88
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/a74f18a1-bc74-491e-b9ac-a365e5f22dcc.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-opus-4-1-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 (August 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "total_cost_usd": "477.65"
30
+ }
31
+ },
32
+ "evaluation_results": [
33
+ {
34
+ "evaluation_name": "SWE-bench Verified Mini",
35
+ "source_data": {
36
+ "source_type": "url",
37
+ "dataset_name": "SWE-bench Verified Mini",
38
+ "url": [
39
+ "https://www.swebench.com",
40
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
41
+ ]
42
+ },
43
+ "metric_config": {
44
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
45
+ "lower_is_better": false,
46
+ "score_type": "continuous",
47
+ "min_score": 0.0,
48
+ "max_score": 1.0
49
+ },
50
+ "score_details": {
51
+ "score": 0.42,
52
+ "details": {
53
+ "accuracy_raw": "42.00%"
54
+ }
55
+ },
56
+ "generation_config": {
57
+ "generation_args": {
58
+ "agentic_eval_config": {
59
+ "available_tools": [
60
+ {
61
+ "name": "bash",
62
+ "description": "Execute shell commands"
63
+ },
64
+ {
65
+ "name": "edit_file",
66
+ "description": "Edit files in the repository"
67
+ },
68
+ {
69
+ "name": "read_file",
70
+ "description": "Read files from the repository"
71
+ }
72
+ ]
73
+ }
74
+ },
75
+ "additional_details": {
76
+ "agent_scaffold": "HAL Generalist Agent",
77
+ "hal_rank": "15",
78
+ "runs": "1",
79
+ "verified": "True",
80
+ "is_pareto": "False",
81
+ "total_cost_usd": "477.65"
82
+ }
83
+ }
84
+ }
85
+ ]
86
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/dd76a080-eb8e-4161-b709-99132ea09733.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/hal-generalist-agent/claude-opus-4-1-high-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 High (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 High (August 2025)",
27
+ "agent_scaffold": "HAL Generalist Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "399.93"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "SWE-bench Verified Mini",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "SWE-bench Verified Mini",
39
+ "url": [
40
+ "https://www.swebench.com",
41
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.46,
53
+ "details": {
54
+ "accuracy_raw": "46.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "edit_file",
67
+ "description": "Edit files in the repository"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the repository"
72
+ }
73
+ ]
74
+ }
75
+ },
76
+ "additional_details": {
77
+ "agent_scaffold": "HAL Generalist Agent",
78
+ "hal_rank": "11",
79
+ "runs": "1",
80
+ "verified": "True",
81
+ "is_pareto": "False",
82
+ "total_cost_usd": "399.93"
83
+ }
84
+ }
85
+ }
86
+ ]
87
+ }
data/hal/hal-swebench-verified-mini/anthropic/claude-opus-4-1/f2746072-cb8f-4efd-9b1a-3dc74be1ee38.json ADDED
@@ -0,0 +1,87 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema_version": "0.2.2",
3
+ "evaluation_id": "hal-swebench-verified-mini/swe-agent/claude-opus-4-1-high-august-2025/1782803417.6616569",
4
+ "retrieved_timestamp": "1782803417.6616569",
5
+ "source_metadata": {
6
+ "source_name": "HAL Leaderboard — SWE-bench Verified Mini",
7
+ "source_type": "documentation",
8
+ "source_organization_name": "Princeton SAgE Team",
9
+ "source_organization_url": "https://hal.cs.princeton.edu",
10
+ "evaluator_relationship": "third_party",
11
+ "additional_details": {
12
+ "paper": "https://arxiv.org/pdf/2510.11977",
13
+ "benchmark_category": "Software Engineering",
14
+ "benchmark_slug": "swebench_verified_mini"
15
+ }
16
+ },
17
+ "eval_library": {
18
+ "name": "HAL",
19
+ "version": "unknown"
20
+ },
21
+ "model_info": {
22
+ "name": "Claude Opus 4.1 High (August 2025)",
23
+ "id": "anthropic/claude-opus-4-1",
24
+ "developer": "anthropic",
25
+ "additional_details": {
26
+ "hal_model_name": "Claude Opus 4.1 High (August 2025)",
27
+ "agent_scaffold": "SWE-Agent",
28
+ "benchmark": "SWE-bench Verified Mini",
29
+ "inference_effort": "high",
30
+ "total_cost_usd": "1599.9"
31
+ }
32
+ },
33
+ "evaluation_results": [
34
+ {
35
+ "evaluation_name": "SWE-bench Verified Mini",
36
+ "source_data": {
37
+ "source_type": "url",
38
+ "dataset_name": "SWE-bench Verified Mini",
39
+ "url": [
40
+ "https://www.swebench.com",
41
+ "https://hal.cs.princeton.edu/swebench_verified_mini"
42
+ ]
43
+ },
44
+ "metric_config": {
45
+ "evaluation_description": "Fraction of 50 verified GitHub issues resolved (0.0–1.0)",
46
+ "lower_is_better": false,
47
+ "score_type": "continuous",
48
+ "min_score": 0.0,
49
+ "max_score": 1.0
50
+ },
51
+ "score_details": {
52
+ "score": 0.54,
53
+ "details": {
54
+ "accuracy_raw": "54.00%"
55
+ }
56
+ },
57
+ "generation_config": {
58
+ "generation_args": {
59
+ "agentic_eval_config": {
60
+ "available_tools": [
61
+ {
62
+ "name": "bash",
63
+ "description": "Execute shell commands"
64
+ },
65
+ {
66
+ "name": "edit_file",
67
+ "description": "Edit files in the repository"
68
+ },
69
+ {
70
+ "name": "read_file",
71
+ "description": "Read files from the repository"
72
+ }
73
+ ]
74
+ }
75
+ },
76
+ "additional_details": {
77
+ "agent_scaffold": "SWE-Agent",
78
+ "hal_rank": "6",
79
+ "runs": "1",
80
+ "verified": "True",
81
+ "is_pareto": "False",
82
+ "total_cost_usd": "1599.9"
83
+ }
84
+ }
85
+ }
86
+ ]
87
+ }