EvidenceDesk
A support copilot that answers only from retrieved documentation, cites verbatim quotes, abstains when the evidence is missing, and can only propose a note that a human must approve. Generation runs on a fine-tuned, 4-bit Gemma 4 12B distilled from Gemma 4 26B-A4B and trained on a laptop.
Loading recorded runs…
Recorded runs
Evaluation
| Cloudflare Llama 3.1 8B | Gemma 4 12B QAT Q4_0 | Fine-tuned Q4_0 | |
|---|---|---|---|
| Regression set, end-to-end pass (30 generations) | 0.800 | 0.867 | 0.900 |
| Regression set p50 / p95 | 1.7 s / 3.1 s (cloud GPU) | 7.1 s / 22.3 s | 2.6 s / 6.1 s |
| Held-out products, pass rate (104, EN + TR) | – | 0.942 | 0.971 |
| Held-out Turkish | – | 0.925 | 0.975 |
| Held-out without grammar-constrained decoding | – | 0.269 | 0.904 |
| Mean output tokens | – | 244 | 109 |
Run it live
This page is a static Space, the only kind a free Hugging Face account can
host, so it shows recorded runs. The interactive ZeroGPU app is in
deploy/hf-space. Locally, the model runs with llama.cpp:
llama-server --hf-repo UmitVice/evidencedesk-gemma-4-12b-GGUF --hf-file evidencedesk-gemma-4-12b-Q4_0.gguf \ --jinja -c 8192 --cache-ram 0 --port 8080 # then point EvidenceDesk at it: GENERATION_PROVIDER=openai_compatible GENERATION_BASE_URL=http://127.0.0.1:8080/v1