Downloads · 30 days
8
20% of all-time downloads
gemma4-31b-rovochat-lora-r64-lr1e4
tzchen07/gemma4-31b-rovochat-lora-r64-lr1e4
gemma4-31b-rovochat-lora-r64-lr1e4 is a machine learning model from tzchen07. Use it for the machine learning task on the model card, and read the license before you ship it in a product. The card lists the license as gemma.
LoRA SFT (rank=64, alpha=128, LR=1e-4, seq=16384, 1 epoch) merged into google/gemma-4-31B-it. One point in a rank x LR sweep on the Rovo Chat Orchestrator tool-calling task.
All-time downloads
40
Public
Parameters
31.3B
62.6 GB on disk
Likes
0
Public
Repo makeup
Click a slice to open those files.
.safetensors62.5 GB · 100%
At a glance
- License
- gemma
- Model type
- gemma4
- Access
- Public
- Created
- Jul 1, 2026
- Updated
- Jul 1, 2026
- SHA
- 8d5fa51a
Base models
Model card
From the Hugging Face model README
Gemma-4-31B Rovo Chat Orchestrator — LoRA r=64 (LR 1e-4), merged
LoRA SFT (rank=64, alpha=128, LR=1e-4, seq=16384, 1 epoch) merged into google/gemma-4-31B-it. One point in a rank x LR sweep on the Rovo Chat Orchestrator tool-calling task.
Eval (725 held-out examples, Claude Opus 4.8 judge)
| metric | this model | base gemma-4-31B-it |
|---|---|---|
| Text AA (shallow) | 82.1% | 96.1% |
| Tool AA (shallow) | 51.0% | 53.2% |
| Overall AA | 73.9% | 84.8% |
| Text AA Deep (binary, 1-5 rubric) | pending% | 86.7% |
| Deep quality mean (1-5) | pending | 4.27 |
| False-tool-trigger rate | 15.1% | 1.9% |
| Runaway rate | 16.4% | 0.3% |
Sweep finding
Across the rank x LR grid, learning rate dominates both accuracy and stability: every LR=5e-5 config beats every LR=1e-4 config on overall AA and runaway rate. Instability is approximately (adapter capacity x LR) = how far the weights travel from the RLHF-aligned base. The sweet spot is r=16, LR=5e-5 (sibling repo tzchen07/gemma4-31b-rovochat-lora-r16-lr5e5).
No fine-tune beats the base -it model overall — narrow-domain tuning erodes the base's tool-vs-text calibration faster than it gains tool accuracy. Base wins on text AA, overall, and stability; the best tune improves only tool AA (+3 pts). Companion full-FT analysis (LR is the killing factor): tzchen07/gemma4-31b-rovochat-sft-v3.
- Type
- gemma4
- License
- gemma
- Created
- Jul 1, 2026
- Updated
- Jul 1, 2026