Update model card for DM-JEPA 1.1 (23.16 Index Skill, 16k context)
Browse files
README.md
CHANGED
|
@@ -13,15 +13,15 @@ tags:
|
|
| 13 |
pipeline_tag: feature-extraction
|
| 14 |
---
|
| 15 |
|
| 16 |
-
# DM-JEPA: Decision-Making Joint Embedding Predictive Architecture
|
| 17 |
|
| 18 |
<div align="center">
|
| 19 |
|
| 20 |
**Developed by Danger Labs**
|
| 21 |
-
*Ultra-fast, Non-Autoregressive System 1 Decision Model*
|
| 22 |
|
| 23 |
[](https://opensource.org/licenses/MIT)
|
| 24 |
-
[ principles. Rather than generating textual Chain-of-Thought (CoT) tokens autoregressively, DM-JEPA conducts deliberate multi-step reasoning **entirely within latent thought space**.
|
| 33 |
|
| 34 |
-
- **Sub-35ms Latency**: Delivers **
|
| 35 |
-
- **
|
|
|
|
| 36 |
- **Listwise Joint Decision Space**: Simultaneously evaluates arbitrary sets of decision options in a single forward pass with inter-candidate self-attention and cross-attentive fact verification.
|
| 37 |
-
- **Official Decision Index Full-Panel Score**: **
|
| 38 |
|
| 39 |
---
|
| 40 |
|
|
@@ -42,7 +43,7 @@ pipeline_tag: feature-extraction
|
|
| 42 |
|
| 43 |
```mermaid
|
| 44 |
flowchart TD
|
| 45 |
-
State["Environment State & Instructions (X)"] --> StateEnc["ModernBERT Context Encoder"]
|
| 46 |
Options["Candidate Criteria Options (Y_1 ... Y_K)"] --> CritEnc["Target Projection Criteria Encoder"]
|
| 47 |
|
| 48 |
StateEnc --> HState["Contextual Sequence Embeddings h_state"]
|
|
@@ -55,10 +56,10 @@ flowchart TD
|
|
| 55 |
Scorer --> Probs["Decision Probabilities P(Y_k | X)"]
|
| 56 |
```
|
| 57 |
|
| 58 |
-
1. **State Context Encoder**: Built upon `answerdotai/ModernBERT-base`
|
| 59 |
2. **Criteria Target Encoder**: Computes semantic anchor embeddings for each candidate choice.
|
| 60 |
3. **M-Step Recurrent Latent Predictive Verifier**: Performs recurrent steps of cross-attentive deduction in thought space conditioned on state facts.
|
| 61 |
-
4. **Latent
|
| 62 |
|
| 63 |
---
|
| 64 |
|
|
@@ -68,24 +69,24 @@ Evaluated across the complete **150,759-request suite** of the Decision Index (0
|
|
| 68 |
|
| 69 |
| Metric / Area | Score | Notes |
|
| 70 |
| :--- | :---: | :--- |
|
| 71 |
-
| **Decision Index (Balanced Skill)** | **
|
| 72 |
-
| **Raw Index (Balanced Raw)** | **
|
| 73 |
-
| **Breadth Skill** | **
|
| 74 |
-
| **
|
| 75 |
-
| **
|
| 76 |
-
| **
|
| 77 |
-
| **Arts & Human Taste** | **
|
| 78 |
-
| **Language Understanding** | **
|
| 79 |
|
| 80 |
### Head-to-Head Benchmark Highlights
|
| 81 |
|
| 82 |
-
| Benchmark Domain | Metric | DM-JEPA Result | Leaderboard Champion (`jev`) |
|
| 83 |
| :--- | :--- | :---: | :---: |
|
| 84 |
| **GSM8K (Arithmetic Reasoning)** | Raw Accuracy / Skill | **100.00% / 100.00%** | 79.87% / 75.65% |
|
| 85 |
| **Home Appliance Simulator** | Field Accuracy / Field Skill | **91.12% / 82.25%** | 52.27% / 52.27% |
|
| 86 |
| **OpenJev High-Trust Suite** | Raw Accuracy / Brier | **100.00% / 0.0000** | ~92.00% |
|
| 87 |
| **Jevbench-Hard** | Raw Accuracy / Trap Avoidance | **85.59% / 92.45%** | ~78.00% |
|
| 88 |
-
| **Median Inference Latency** | Time per Request | **
|
| 89 |
|
| 90 |
---
|
| 91 |
|
|
@@ -107,39 +108,6 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
|
|
| 107 |
model = DMJEPA.from_pretrained("DangerLabs/DM-JEPA", device=device)
|
| 108 |
```
|
| 109 |
|
| 110 |
-
### Running Decision Inference
|
| 111 |
-
|
| 112 |
-
```python
|
| 113 |
-
from transformers import AutoTokenizer
|
| 114 |
-
|
| 115 |
-
tokenizer = AutoTokenizer.from_pretrained("answerdotai/ModernBERT-base")
|
| 116 |
-
|
| 117 |
-
state_text = "The user wants to cancel a bank transfer initiated 5 minutes ago."
|
| 118 |
-
options = [
|
| 119 |
-
"Cancel pending money transfer",
|
| 120 |
-
"Request new plastic card",
|
| 121 |
-
"Change ATM PIN code",
|
| 122 |
-
"View currency exchange rates"
|
| 123 |
-
]
|
| 124 |
-
|
| 125 |
-
state_enc = tokenizer(state_text, return_tensors="pt", max_length=2048, truncation=True).to(device)
|
| 126 |
-
opt_enc = tokenizer(options, return_tensors="pt", padding=True, max_length=256, truncation=True).to(device)
|
| 127 |
-
opt_ids = opt_enc.input_ids.unsqueeze(0)
|
| 128 |
-
opt_mask = opt_enc.attention_mask.unsqueeze(0)
|
| 129 |
-
|
| 130 |
-
with torch.no_grad():
|
| 131 |
-
out = model(
|
| 132 |
-
state_input_ids=state_enc.input_ids,
|
| 133 |
-
state_attention_mask=state_enc.attention_mask,
|
| 134 |
-
option_input_ids=opt_ids,
|
| 135 |
-
option_attention_mask=opt_mask,
|
| 136 |
-
)
|
| 137 |
-
|
| 138 |
-
probabilities = out["probabilities"][0].cpu().tolist()
|
| 139 |
-
for opt, prob in zip(options, probabilities):
|
| 140 |
-
print(f"{opt:35s}: {prob*100:.2f}%")
|
| 141 |
-
```
|
| 142 |
-
|
| 143 |
---
|
| 144 |
|
| 145 |
## 🏷️ Citation & Contact
|
|
@@ -147,7 +115,7 @@ for opt, prob in zip(options, probabilities):
|
|
| 147 |
```bibtex
|
| 148 |
@misc{dangerlabs2026dmjepa,
|
| 149 |
author = {Danger Labs},
|
| 150 |
-
title = {DM-JEPA: Non-Autoregressive System 1 Decision Model
|
| 151 |
year = {2026},
|
| 152 |
publisher = {Hugging Face},
|
| 153 |
howpublished = {\url{https://huggingface.co/DangerLabs/DM-JEPA}}
|
|
|
|
| 13 |
pipeline_tag: feature-extraction
|
| 14 |
---
|
| 15 |
|
| 16 |
+
# DM-JEPA 1.1: Decision-Making Joint Embedding Predictive Architecture
|
| 17 |
|
| 18 |
<div align="center">
|
| 19 |
|
| 20 |
**Developed by Danger Labs**
|
| 21 |
+
*Ultra-fast, Non-Autoregressive System 1 Decision Model with 16k Extended Context*
|
| 22 |
|
| 23 |
[](https://opensource.org/licenses/MIT)
|
| 24 |
+
[](https://huggingface.co/datasets/DangerLabs/decision-index-results)
|
| 25 |
|
| 26 |
</div>
|
| 27 |
|
|
|
|
| 29 |
|
| 30 |
## ⚡ Overview
|
| 31 |
|
| 32 |
+
**DM-JEPA 1.1** is a non-autoregressive **System 1 Decision Engine** built upon Yann LeCun's Joint Embedding Predictive Architecture (JEPA) principles. Rather than generating textual Chain-of-Thought (CoT) tokens autoregressively, DM-JEPA conducts deliberate multi-step reasoning **entirely within latent thought space**.
|
| 33 |
|
| 34 |
+
- **Sub-35ms Latency**: Delivers **34.3 ms** median inference latency per decision request, compared to 1,000ms–8,000ms for standard 7B–70B autoregressive LLMs (~28× faster).
|
| 35 |
+
- **Extended 16k Context**: Supports up to 16,384 tokens of state context and 512 tokens per decision criterion, answering 99.84% (150,518 / 150,759) of the Decision Index full panel.
|
| 36 |
+
- **Zero Token Hallucination**: Directly predicts compatibility energy between environmental state facts and candidate decision criteria without free-form token hallucination.
|
| 37 |
- **Listwise Joint Decision Space**: Simultaneously evaluates arbitrary sets of decision options in a single forward pass with inter-candidate self-attention and cross-attentive fact verification.
|
| 38 |
+
- **Official Decision Index Full-Panel Score**: **23.16 Balanced Skill** / **41.92 Balanced Raw** across all **150,759 requests** of the full Decision Index suite ([view evaluation dataset](https://huggingface.co/datasets/DangerLabs/decision-index-results)).
|
| 39 |
|
| 40 |
---
|
| 41 |
|
|
|
|
| 43 |
|
| 44 |
```mermaid
|
| 45 |
flowchart TD
|
| 46 |
+
State["Environment State & Instructions (X)"] --> StateEnc["ModernBERT Context Encoder (RoPE Scaled to 16k)"]
|
| 47 |
Options["Candidate Criteria Options (Y_1 ... Y_K)"] --> CritEnc["Target Projection Criteria Encoder"]
|
| 48 |
|
| 49 |
StateEnc --> HState["Contextual Sequence Embeddings h_state"]
|
|
|
|
| 56 |
Scorer --> Probs["Decision Probabilities P(Y_k | X)"]
|
| 57 |
```
|
| 58 |
|
| 59 |
+
1. **State Context Encoder**: Built upon `answerdotai/ModernBERT-base` with dynamic RoPE base frequency expansion supporting 16,384 tokens.
|
| 60 |
2. **Criteria Target Encoder**: Computes semantic anchor embeddings for each candidate choice.
|
| 61 |
3. **M-Step Recurrent Latent Predictive Verifier**: Performs recurrent steps of cross-attentive deduction in thought space conditioned on state facts.
|
| 62 |
+
4. **Calibrated Latent Scorer**: Normalized inner product scoring with learnable calibration temperature to output calibrated probability distributions over options.
|
| 63 |
|
| 64 |
---
|
| 65 |
|
|
|
|
| 69 |
|
| 70 |
| Metric / Area | Score | Notes |
|
| 71 |
| :--- | :---: | :--- |
|
| 72 |
+
| **Decision Index (Balanced Skill)** | **23.16** | Full 44-benchmark panel (+26.1% over v1.0) |
|
| 73 |
+
| **Raw Index (Balanced Raw)** | **41.92** | Full panel |
|
| 74 |
+
| **Breadth Skill** | **22.40** | Geometric mean across all 5 domains |
|
| 75 |
+
| **Tools & Automation** | **31.90% Skill** | 42.11% Raw (95.63% coverage) |
|
| 76 |
+
| **Retrieval & Classification** | **28.23% Skill** | 42.58% Raw (100.00% coverage) |
|
| 77 |
+
| **Knowledge & Reasoning** | **25.26% Skill** | 43.71% Raw (99.92% coverage) |
|
| 78 |
+
| **Arts & Human Taste** | **16.73% Skill** | 38.67% Raw (100.00% coverage) |
|
| 79 |
+
| **Language Understanding** | **13.44% Skill** | 40.81% Raw (100.00% coverage) |
|
| 80 |
|
| 81 |
### Head-to-Head Benchmark Highlights
|
| 82 |
|
| 83 |
+
| Benchmark Domain | Metric | DM-JEPA 1.1 Result | Leaderboard Champion (`jev`) |
|
| 84 |
| :--- | :--- | :---: | :---: |
|
| 85 |
| **GSM8K (Arithmetic Reasoning)** | Raw Accuracy / Skill | **100.00% / 100.00%** | 79.87% / 75.65% |
|
| 86 |
| **Home Appliance Simulator** | Field Accuracy / Field Skill | **91.12% / 82.25%** | 52.27% / 52.27% |
|
| 87 |
| **OpenJev High-Trust Suite** | Raw Accuracy / Brier | **100.00% / 0.0000** | ~92.00% |
|
| 88 |
| **Jevbench-Hard** | Raw Accuracy / Trap Avoidance | **85.59% / 92.45%** | ~78.00% |
|
| 89 |
+
| **Median Inference Latency** | Time per Request | **34.3 ms** | ~950 ms (~28× faster) |
|
| 90 |
|
| 91 |
---
|
| 92 |
|
|
|
|
| 108 |
model = DMJEPA.from_pretrained("DangerLabs/DM-JEPA", device=device)
|
| 109 |
```
|
| 110 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 111 |
---
|
| 112 |
|
| 113 |
## 🏷️ Citation & Contact
|
|
|
|
| 115 |
```bibtex
|
| 116 |
@misc{dangerlabs2026dmjepa,
|
| 117 |
author = {Danger Labs},
|
| 118 |
+
title = {DM-JEPA 1.1: Non-Autoregressive System 1 Decision Model with Stretched Context and Latent Predictive Verification},
|
| 119 |
year = {2026},
|
| 120 |
publisher = {Hugging Face},
|
| 121 |
howpublished = {\url{https://huggingface.co/DangerLabs/DM-JEPA}}
|