CoolFace
Modelpublic

Ailiance-fr/router-v5-minilm

sourceHugging Faceapache-2.0updated 5mo agoView on Hugging Face
0likes
Model Card

ailiance-router-v5 (MiniLM L6 v2 + 2-layer MLP)

Domain classifier head used by the ailiance gateway to route incoming chat requests to the most adapted backend worker.

Performance

  • —Top-1 accuracy: 87.6 %
  • —Top-3 accuracy: 98.7 %
  • —Validation set: 1 978 prompts across 32 domains.

Architecture

  • —Encoder: sentence-transformers/all-MiniLM-L6-v2 (frozen, 384d)
  • —Head: 2-layer MLP Linear(384→256) -> GELU -> Dropout(0.1) -> Linear(256→32)
  • —Loss: BCE-with-logits (multi-label sigmoid)
  • —Optimizer: AdamW lr=1e-3, batch=128, 30 epochs, seed=42

🇪🇺 EU AI Act transparency

This artefact is published under the AI Act framework (Regulation EU 2024/1689).

FieldValue
ProviderAiliance (clemsail)
Role under AI ActGPAI provider (this head). Encoder is upstream Microsoft / sentence-transformers under Apache-2.0.
Adapter typeCustom 2-layer MLP head (~85 k parameters)
Base modelsentence-transformers/all-MiniLM-L6-v2
LicenseApache-2.0
Intended useDomain routing only — classify a user prompt over 32 technical domains, decide which downstream worker handles the request. Not for standalone classification of safety-critical text.
Out of scopeHealthcare triage, legal classification, autonomous safety decisions, anything requiring calibrated probabilities.
Risk classificationLimited risk — Article 50 obligations apply.
Training data originAll open permissive-licensed datasets + Ailiance in-house curation. Full per-source provenance: https://ailiance.fr/transparency/router-provenance
Reproducibilityhttps://ailiance.fr/transparency/router-rebuild-recipe
Contactpostmaster@saillant.cc

Training data sources

Domain groupSource datasetLicenseRows
python, rust, ts, cpp, shell, html-css, security, devopssahil2801/CodeAlpaca-20kCC-BY-4.0~3 700
sqlb-mc2/sql-create-contextCC-BY-4.0805
chat-frOpenAssistant/oasst1 (FR-filtered)Apache-2.0801
mathopenai/gsm8kMIT803
reasoningAI-MO/NuminaMath-CoTApache-2.0803
spice, power, electronicstheprint/Electronics-QAApache-2.0~1 500
freecadYas1n/FreeCAD_Sketches (filename-derived)CC-BY-4.0602
lua-upyRoblox/luau_corpusMIT603
calcul-normatif, docker, FR/EN augment, greetingsAiliance in-houseApache-2.0528 + 200

Total clean corpus: 9 817 prompts across 32 domains.

Files

  • —router.safetensors — MLP head weights
  • —meta.json — encoder name, dim, head config, domain list
  • —label_map.json — domain ↔ index

Validated in ailiance/ailiance-bench v0.2

This model is referenced in the Ailiance benchmark suite (Phase 6 scoreboard, 7-task hardware-design evaluation).

See the full scoreboard: ailiance-bench README#scoreboard-lora-phase-6.