Text Classification
cortiq
Laya
English
cmf
jev
decision-making
structured-output
tool-selection
semantic-routing
on-device
metal
vulkan
intent-classification
resonance-routing
Instructions to use infosave/cmf-decision with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- cortiq
How to use infosave/cmf-decision with cortiq:
# one Rust binary, no additional dependencies cargo install cortiq-cli # or a prebuilt binary from github.com/infosave2007/cmf/releases hf download infosave/cmf-decision --include "*.cmf" --local-dir . ls *.cmf # some repos ship more than one quantization
cortiq run FILE.cmf --prompt "What is the capital of France?"
cortiq serve FILE.cmf --port 8080 # OpenAI-compatible server
- Laya
How to use infosave/cmf-decision with Laya:
# No code snippets available yet for this library. # To use this model, check the repository files and the library's documentation. # Want to help? PRs adding snippets are welcome at: # https://github.com/huggingface/huggingface.js
- Notebooks
- Google Colab
- Kaggle
Download evidence/comparison.json from infosave/cmf-decision: direct link, hf CLI and curl.
- Browser
- Download file 4.82 kB
-
https://huggingface.co/infosave/cmf-decision/resolve/main/evidence/comparison.json
- Command line
-
hf download hf://infosave/cmf-decision/evidence/comparison.json
-
curl -L -o comparison.json https://huggingface.co/infosave/cmf-decision/resolve/main/evidence/comparison.json
4.82 kB
| { | |
| "schema": "cortiq-model-card-comparison/1", | |
| "model_file_sha256": "ed9b8ec2bbfe9e9fd30f14a5eaf82314f38bc7e7510a39772baa2de3801d79b1", | |
| "model_bytes": 304520292, | |
| "runtime": "cortiq 0.7.8", | |
| "oracle_in_hybrid_benchmark": "deepseek/deepseek-v4.1-flash", | |
| "benchmarks": [ | |
| { | |
| "id": "banking77", | |
| "name": "BANKING77", | |
| "n": 3080, | |
| "labels": 77, | |
| "local_correct": 2875, | |
| "local_accuracy_pct": 93.34, | |
| "jev_correct": 2636, | |
| "jev_accuracy_pct": 85.58, | |
| "local_wilson95_pct": [ | |
| 92.41, | |
| 94.17 | |
| ], | |
| "jev_wilson95_pct": [ | |
| 84.3, | |
| 86.78 | |
| ], | |
| "mcnemar_p": 9.72e-33, | |
| "accepted": 2791, | |
| "accepted_correct": 2714, | |
| "accepted_accuracy_pct": 97.24, | |
| "coverage_pct": 90.62, | |
| "local_p50_ms": 3.92, | |
| "local_p95_ms": 5.27, | |
| "cortiq_https_p50_ms": 78.4, | |
| "cortiq_https_p95_ms": 141.6, | |
| "jev_http_p50_ms": 348.33, | |
| "jev_http_p95_ms": 426.03, | |
| "jev_latency_n": 3080, | |
| "jev_latency_concurrency": 4, | |
| "hybrid_correct": 2893, | |
| "hybrid_accuracy_pct": 93.93, | |
| "hybrid_oracle_calls": 282, | |
| "hybrid_api_usd_per_million": 3.01, | |
| "jev_api_usd_per_million": 183.69 | |
| }, | |
| { | |
| "id": "clinc150", | |
| "name": "CLINC150", | |
| "n": 4500, | |
| "labels": 150, | |
| "local_correct": 4328, | |
| "local_accuracy_pct": 96.18, | |
| "jev_correct": 4354, | |
| "jev_accuracy_pct": 96.76, | |
| "local_wilson95_pct": [ | |
| 95.58, | |
| 96.7 | |
| ], | |
| "jev_wilson95_pct": [ | |
| 96.2, | |
| 97.23 | |
| ], | |
| "mcnemar_p": 0.099, | |
| "accepted": 4145, | |
| "accepted_correct": 4091, | |
| "accepted_accuracy_pct": 98.7, | |
| "coverage_pct": 92.11, | |
| "local_p50_ms": 3.79, | |
| "local_p95_ms": 4.57, | |
| "cortiq_https_p50_ms": 78.5, | |
| "cortiq_https_p95_ms": 134.6, | |
| "jev_http_p50_ms": 391.51, | |
| "jev_http_p95_ms": 528.49, | |
| "jev_latency_n": 200, | |
| "jev_latency_concurrency": 1, | |
| "hybrid_correct": 4386, | |
| "hybrid_accuracy_pct": 97.47, | |
| "hybrid_oracle_calls": 351, | |
| "hybrid_api_usd_per_million": 3.53, | |
| "jev_api_usd_per_million": 271.61 | |
| }, | |
| { | |
| "id": "massive", | |
| "name": "MASSIVE", | |
| "n": 2974, | |
| "labels": 60, | |
| "local_correct": 2562, | |
| "local_accuracy_pct": 86.15, | |
| "jev_correct": 2551, | |
| "jev_accuracy_pct": 85.78, | |
| "local_wilson95_pct": [ | |
| 84.86, | |
| 87.34 | |
| ], | |
| "jev_wilson95_pct": [ | |
| 84.48, | |
| 86.99 | |
| ], | |
| "mcnemar_p": 0.607, | |
| "accepted": 1615, | |
| "accepted_correct": 1581, | |
| "accepted_accuracy_pct": 97.89, | |
| "coverage_pct": 54.3, | |
| "local_p50_ms": 3.0, | |
| "local_p95_ms": 3.58, | |
| "cortiq_https_p50_ms": 70.4, | |
| "cortiq_https_p95_ms": 117.9, | |
| "jev_http_p50_ms": 346.37, | |
| "jev_http_p95_ms": 492.93, | |
| "jev_latency_n": 200, | |
| "jev_latency_concurrency": 1, | |
| "hybrid_correct": 2617, | |
| "hybrid_accuracy_pct": 88.0, | |
| "hybrid_oracle_calls": 1335, | |
| "hybrid_api_usd_per_million": 8.48, | |
| "jev_api_usd_per_million": 110.86 | |
| } | |
| ], | |
| "notes": [ | |
| "Local accuracy is top-1 over every test row before abstention; gate accuracy and coverage are separate.", | |
| "Local CPU latency: full text to decision, Apple M4, one CPU thread, 50 warmups, all test rows.", | |
| "Cortiq HTTPS latency: same client Mac to deployed server, sequential, no oracle, all test rows.", | |
| "Jev latency includes network; BANKING77 uses four workers, CLINC150 and MASSIVE serial samples of 200.", | |
| "Different hardware and concurrency: this is not a hardware-normalized throughput comparison.", | |
| "Hybrid costs: served cache/learning run with stored DeepSeek answers replayed and 24 live MASSIVE calls; not a timing run.", | |
| "Per-million fees extrapolate recorded costs, excluding hardware, electricity and operations.", | |
| "Reused public benchmarks, unequal supervision: CMF train+dev vs two examples per label for Jev.", | |
| "Current model uses BAAI bge-small-en-v1.5 compressed by Cortiq NVG, not the separate Embryo experiment." | |
| ], | |
| "sources": [ | |
| { | |
| "path_in_project": "artifacts/decision-v4-20260926/gates-release.json", | |
| "sha256": "47d3039559e0f87dee0c6dae9adc3bbd63acc52b05d4859f2de82cf7a0a9ba03" | |
| }, | |
| { | |
| "path_in_project": "reports/decision-v4-20260926/live/live.json", | |
| "sha256": "c8769af100e3138291675ab3eed20caa3ecaa780b6b81f4cecbde96dc29f87f5" | |
| }, | |
| { | |
| "path_in_project": "reports/decision-v4-20260926/article/stored.json", | |
| "sha256": "4d6918a2d98f16c8b4cdda7759fef98e6343f091fa7723e8ea938d9fb696db47" | |
| }, | |
| { | |
| "path_in_project": "reports/decision-v4-20260926/eval/stream-final.json", | |
| "sha256": "0bb19de37cee315a5a5ad4563abcf2944a80c49ef05984554a4fc31012702f50" | |
| } | |
| ] | |
| } | |