Skip to content

Feat/claude opus 4 7 #5

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
Khurdhula-Harshavardhan wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Feat/claude opus 4 7 #5

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
ae76ac0
feat(anthropic): support Claude Opus 4.7 by omitting deprecated tempe…
Khurdhula-Harshavardhan Apr 30, 2026
4fcfa42
chore(deps): make vLLM an optional extra so base install works on macOS
Khurdhula-Harshavardhan Apr 30, 2026
d239982
Merge feat/opus-4-7-support: omit deprecated temperature for Opus 4.7
Khurdhula-Harshavardhan Apr 30, 2026
02b38cf
feat(opus-4.7): add scores for opus 4.7
Khurdhula-Harshavardhan Apr 30, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -98,6 +98,12 @@ make lint # ruff check .

For local vLLM inference (NVIDIA GPU, CUDA 12.8, ≥ 24 GB VRAM):

```bash
uv sync --extra vllm
```

vLLM is an optional extra so the base install resolves on macOS / non-CUDA hosts. If you don't use `uv`:

```bash
uv pip install vllm --extra-index-url https://download.pytorch.org/whl/cu128
```
Expand Down
115 changes: 115 additions & 0 deletions data/audio_responses/response_claude-opus-4-7_audio.jsonl
View file
Open in desktop

Large diffs are not rendered by default.

115 changes: 115 additions & 0 deletions data/evaluation/audio/claude-opus-4-7/eval_records.jsonl
View file
Open in desktop

Large diffs are not rendered by default.

264 changes: 264 additions & 0 deletions data/evaluation/audio/claude-opus-4-7/eval_summary.json
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,264 @@
{
"response_file": "data/audio_responses/response_claude-opus-4-7_audio.jsonl",
"num_records": 115,
"model_ids": [
"claude-opus-4-7"
],
"data_quality": {
"json_parse_fail_count": 16,
"json_non_structured_root_count": 16,
"invalid_schema_input_count": 0,
"unknown_difficulty_count": 0,
"malformed_jsonl_line_count": 0
},
"summary": {
"overall": {
"n": 115,
"metrics": {
"json_parse_success": {
"mean": 0.8608695652173913,
"ci95_low": 0.7913043478260869,
"ci95_high": 0.9217391304347826,
"metric_name": "JSON Parse Success"
},
"json_root_structured": {
"mean": 0.8608695652173913,
"ci95_low": 0.7913043478260869,
"ci95_high": 0.9217391304347826,
"metric_name": "Structured JSON Root"
},
"schema_valid_input": {
"mean": 1.0,
"ci95_low": 1.0,
"ci95_high": 1.0,
"metric_name": "Schema Valid Input"
},
"schema_compliance": {
"mean": 0.8434782608695652,
"ci95_low": 0.7739130434782608,
"ci95_high": 0.9043478260869565,
"metric_name": "JSON Pass Rate"
},
"leaf_value_em": {
"mean": 0.19118691205824562,
"ci95_low": 0.1580874948383292,
"ci95_high": 0.2244665723935802,
"metric_name": "Truth Score"
},
"value_token_f1": {
"mean": 0.39953452516147003,
"ci95_low": 0.354763558268282,
"ci95_high": 0.44584762494411256,
"metric_name": "Faithfulness Score"
},
"hier_path_recall": {
"mean": 0.8099408675778713,
"ci95_low": 0.7375096396529655,
"ci95_high": 0.8672628381042644,
"metric_name": "Path Recall"
},
"path_set_f1": {
"mean": 0.7567576033226274,
"ci95_low": 0.694100517443477,
"ci95_high": 0.817306853440761,
"metric_name": "Structure Coverage"
},
"type_precision": {
"mean": 0.8434782608695652,
"ci95_low": 0.7739130434782608,
"ci95_high": 0.9043478260869565,
"metric_name": "Type Safety"
},
"strict_json_em": {
"mean": 0.0,
"ci95_low": 0.0,
"ci95_high": 0.0,
"metric_name": "Perfect Response Rate"
}
},
"category_scores": {
"Long Context Extraction": {
"mean": 0.466887434932529,
"ci95_low": 0.4240569771005347,
"ci95_high": 0.5078234429530224,
"category_name": "Long Context Extraction",
"components": [
"leaf_value_em",
"value_token_f1",
"hier_path_recall"
]
},
"Complex Schema Handling": {
"mean": 0.814571375020586,
"ci95_low": 0.7468215705356411,
"ci95_high": 0.8763733343593071,
"category_name": "Complex Schema Handling",
"components": [
"schema_compliance",
"path_set_f1",
"type_precision"
]
},
"Multi-Context Linking": {
"mean": 0.2953607186098578,
"ci95_low": 0.25861154855105634,
"ci95_high": 0.3337783826129957,
"category_name": "Multi-Context Linking",
"components": [
"leaf_value_em",
"value_token_f1"
]
},
"Output Contract Reliability": {
"mean": 0.8492753623188406,
"ci95_low": 0.7855072463768116,
"ci95_high": 0.9101449275362319,
"category_name": "Output Contract Reliability",
"components": [
"json_parse_success",
"schema_compliance",
"type_precision"
]
},
"Strict Precision": {
"mean": 0.0,
"ci95_low": 0.0,
"ci95_high": 0.0,
"category_name": "Strict Precision",
"components": [
"strict_json_em"
]
}
}
},
"overall_weighted": {
"n": 115,
"metrics": {
"json_parse_success": {
"mean": 0.8600583090379009,
"ci95_low": 0.7906976744186046,
"ci95_high": 0.9294117647058824,
"metric_name": "JSON Parse Success"
},
"json_root_structured": {
"mean": 0.8600583090379009,
"ci95_low": 0.7982456140350878,
"ci95_high": 0.9212827988338192,
"metric_name": "Structured JSON Root"
},
"schema_valid_input": {
"mean": 1.0,
"ci95_low": 1.0,
"ci95_high": 1.0,
"metric_name": "Schema Valid Input"
},
"schema_compliance": {
"mean": 0.8425655976676385,
"ci95_low": 0.7719298245614035,
"ci95_high": 0.9125364431486881,
"metric_name": "JSON Pass Rate"
},
"leaf_value_em": {
"mean": 0.19044641697880788,
"ci95_low": 0.16004373370483765,
"ci95_high": 0.22128222987239432,
"metric_name": "Truth Score"
},
"value_token_f1": {
"mean": 0.39898713808088415,
"ci95_low": 0.35519676481217377,
"ci95_high": 0.445811415929564,
"metric_name": "Faithfulness Score"
},
"hier_path_recall": {
"mean": 0.8088326510622903,
"ci95_low": 0.7401982594314398,
"ci95_high": 0.8709379549287983,
"metric_name": "Path Recall"
},
"path_set_f1": {
"mean": 0.7560357493737474,
"ci95_low": 0.69079495117598,
"ci95_high": 0.8150754189782181,
"metric_name": "Structure Coverage"
},
"type_precision": {
"mean": 0.8425655976676385,
"ci95_low": 0.7732558139534884,
"ci95_high": 0.9035087719298246,
"metric_name": "Type Safety"
},
"strict_json_em": {
"mean": 0.0,
"ci95_low": 0.0,
"ci95_high": 0.0,
"metric_name": "Perfect Response Rate"
}
},
"category_scores": {
"Long Context Extraction": {
"mean": 0.46608873537399415,
"ci95_low": 0.42160904615108913,
"ci95_high": 0.5093431833999407,
"category_name": "Long Context Extraction",
"components": [
"leaf_value_em",
"value_token_f1",
"hier_path_recall"
]
},
"Complex Schema Handling": {
"mean": 0.8137223149030081,
"ci95_low": 0.7508078837811929,
"ci95_high": 0.8795349085857703,
"category_name": "Complex Schema Handling",
"components": [
"schema_compliance",
"path_set_f1",
"type_precision"
]
},
"Multi-Context Linking": {
"mean": 0.29471677752984604,
"ci95_low": 0.2574922501008539,
"ci95_high": 0.33127480545154314,
"category_name": "Multi-Context Linking",
"components": [
"leaf_value_em",
"value_token_f1"
]
},
"Output Contract Reliability": {
"mean": 0.8483965014577259,
"ci95_low": 0.7790697674418605,
"ci95_high": 0.9096209912536443,
"category_name": "Output Contract Reliability",
"components": [
"json_parse_success",
"schema_compliance",
"type_precision"
]
},
"Strict Precision": {
"mean": 0.0,
"ci95_low": 0.0,
"ci95_high": 0.0,
"category_name": "Strict Precision",
"components": [
"strict_json_em"
]
}
},
"weighting": "schema_complexity",
"weight_field_priority": [
"schema_complexity",
"difficulty"
],
"difficulty_weights": {
"easy": 1.0,
"medium": 2.0,
"hard": 3.0
}
}
}
}
3 changes: 2 additions & 1 deletion data/evaluation/display_names.json
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,5 +20,6 @@
"gpt-oss": "GPT-OSS-20B",
"inference-net-Schematron-8B": "Schematron-8B",
"ibm-granite-4.0-h-small": "IBM-Granite-4.0",
"interfaze-beta": "Interfaze-Beta"
"interfaze-beta": "Interfaze-Beta",
"claude-opus-4-7": "Claude-Opus-4.7"
}
209 changes: 209 additions & 0 deletions data/evaluation/image/claude-opus-4-7/eval_records.jsonl
View file
Open in desktop

Large diffs are not rendered by default.

Loading
Loading