File size: 5,331 Bytes
f9edec6
f4602c4
 
 
2d0db8d
 
f4602c4
2d0db8d
 
 
 
 
 
 
 
 
 
 
 
f9edec6
2d0db8d
d6cdabf
2294a8a
 
 
 
 
d6cdabf
 
 
 
 
2d0db8d
 
 
 
 
 
d6cdabf
 
 
2d0db8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d6cdabf
 
 
 
 
 
2d0db8d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
d6cdabf
 
 
 
 
 
 
 
 
 
 
2d0db8d
 
 
d6cdabf
2d0db8d
 
f4602c4
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
---
license: other
license_name: mit-and-apache-2.0
license_link: LICENSE
library_name: transformers
base_model: google/gemma-4-26B-A4B-it
base_model_relation: finetune
pipeline_tag: text-classification
language:
- en
tags:
- decision-model
- classification
- calibration
- routing
- guardrails
- vision
- multimodal
- baseten
---

<div align="center">
  <img
    src="https://huggingface.co/baselabs/groundhog-1/resolve/main/assets/banner.png"
    alt="Baseten groundhog-1 decision model"
    style="width: 100%; max-width: 100%; height: auto; display: inline-block; margin-bottom: 0.5em; margin-top: 0.5em;"
  />
  <div style="display: flex; justify-content: center; gap: 0.5em; margin-bottom: 1em;">
    <a href="https://www.baseten.co/library/groundhog-1/"><strong>Try on Baseten</strong></a>
  </div>
</div>

# groundhog-1

groundhog-1 is a 26B parameter mixture-of-experts **decision model** (about 4B active parameters per token) post-trained from [Gemma-4-26B-A4B](https://huggingface.co/google/gemma-4-26B-A4B-it) by Base Labs at Baseten.
Provide the model a state (text, optionally with images) and a set of questions and it will return a calibrated probability for every option,
answering all questions from **one shared-prefix prefill**, without generating text.

- **Long documents at a small-model price**: trained on sequences up to 131,072 tokens and priced like elephant-1.
- **Same training data and recipe family as bison-1**, the #1 decision model on the Artificial Analysis Decision Index v2.1.

For the highest accuracy, see the dense 31B sibling [bison-1](https://huggingface.co/baselabs/bison-1).

## 🗒️ Model Details

| Model | Parameters | Description |
|---|---|---|
| [Gemma-4-26B-A4B](https://huggingface.co/google/gemma-4-26B-A4B-it) | 26B (about 4B active) | General-purpose base model |
| **[groundhog-1](https://huggingface.co/baselabs/groundhog-1)** | 26B (about 4B active) | Post-trained for single-pass, calibrated decisions |
| [bison-1](https://huggingface.co/baselabs/bison-1) | 31B | Dense sibling, highest accuracy |

- **Inputs**: text and images
- **Training**: full fine-tune of the language model on text and image decisions; the vision encoder, expert router, norms and embeddings are frozen
- **Attention**: block-bidirectional mask on full-attention layers; sliding-window layers stay causal.
- **Max sequence length**: 131,072 tokens

## 🏃 How to use

### Baseten Model APIs

groundhog-1 is available on [Baseten Model APIs](https://www.baseten.co/library/groundhog-1/) as `decision-models/baseten-groundhog-latest`, at $0.035 per 1M tokens. It uses the same API as mammoth-1 and elephant-1, and the endpoint accepts text and image inputs.

### Request format

A request carries one `state` and any number of `questions`. Every question is answered in the same call.

```python
questions = {
    "refund": {"type": "noul", "instructions": "Is the customer asking for a refund?"},
    "team": {
        "type": "choice",
        "instructions": "Which team should handle this?",
        "criteria": {"billing": "Charges, refunds, invoices",
                     "technical": "App or site faults",
                     "fraud": "Suspected unauthorised use"},
    },
    "urgency": {
        "type": "score",
        "instructions": "How urgent is this?",
        "criteria": ["Can wait", "Today", "Blocking the customer now"],
    },
}
request = {"state": "I was charged twice this month, please refund one of them.", "questions": questions}
```

| `type` | `criteria` | answer fields |
|---|---|---|
| `noul`: yes or no | optional | `noul`: P(yes) |
| `choice`: one of named options | `{name: description}` | `choice`, `probabilities` |
| `score`: ordered levels | list of levels, lowest first | `score`, `probabilities` |

Requests longer than the trained length limits are refused, not truncated.

## 📊 Performance

### Long context (our run)

| Benchmark | groundhog-1 |
|---|---:|
| LongMemEval M, 128k tokens | 77.4 |
| LongMemEval S, 128k tokens | 79.3 |

groundhog-1 has not yet been scored on the Artificial Analysis Decision Index.

## 🔧 Training

- **Data**: 6.71M decision examples (Mix 12): core decision skills, safety and abstention, judging and expert domains, document and long-context tasks, and public decision data, including public benchmark sets.
- **Objective**: cross-entropy plus a Brier calibration term (λ = 1), with a trained "I don't know" option. About one example in seven carries a probability target instead of a single right answer.
- **Recipe**: 1 epoch (5,621 steps), peak LR 3.95e-6, no weight decay, 3% warm-up, bf16, sequences up to 131,072 tokens.
- **Hardware**: 16 nodes × 8 B200 (128 GPUs).

## License

groundhog-1 is fine-tuned by Baseten from [Gemma 4 26B A4B IT](https://huggingface.co/google/gemma-4-26B-A4B-it), which is licensed under the [Apache License 2.0](https://www.apache.org/licenses/LICENSE-2.0). Baseten's modifications are released under the MIT License. Use of the model is subject to both licenses; see [LICENSE](LICENSE) and [NOTICE](NOTICE).

## Attribution

Base model: Gemma 4 26B A4B IT, made available by Google LLC, licensed under the Apache License 2.0. Baseten modified the weights by further training. Gemma is a trademark of Google LLC. This model is not affiliated with or endorsed by Google.