Pulse Decide 150M v0.1 — weights, config, demo GIF

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+pulse_beat.gif filter=lfs diff=lfs merge=lfs -text

LICENSE

@@ -0,0 +1,40 @@
+Pulse Research Preview License (pulse-research-preview)
+Applies to: Pulse Decide 150M
+Copyright (c) 2026 RouterML
+
+STATUS
+This is an early research preview, not a production release. It is provided for research,
+evaluation and experimentation.
+
+GRANT
+Subject to the conditions below, you are granted a non-exclusive, worldwide, royalty-free
+licence to download, use, evaluate, study and modify these model weights, and to publish
+results obtained with them.
+
+CONDITIONS
+1. Attribution. Retain this notice and credit "Pulse Decide 150M (RouterML)" in any
+   publication, product or derivative that uses these weights.
+
+2. Upstream data licences. These weights were derived from a corpus assembled from ~40
+   public text-classification datasets whose licences differ. Some are permissive; some are
+   research/non-commercial only; some derive from content whose platform Terms of Service
+   restrict redistribution. RouterML publishes the weights, not the training data, and makes
+   no representation that any particular downstream use is permitted by those upstream
+   licences. YOU ARE RESPONSIBLE for verifying that your intended use — in particular any
+   commercial use — complies with the licences and terms of the underlying datasets.
+
+3. No high-stakes use. Not evaluated for fairness, safety or robustness. Do not use for
+   consequential decisions about people (employment, credit, housing, insurance, legal,
+   medical or safety-critical matters) without your own independent evaluation and
+   appropriate human oversight.
+
+4. No endorsement. Do not use RouterML's names or marks to imply endorsement of your work.
+
+DISCLAIMER
+THE WEIGHTS ARE PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
+INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR
+PURPOSE AND NON-INFRINGEMENT. IN NO EVENT SHALL ROUTERML BE LIABLE FOR ANY CLAIM, DAMAGES OR
+OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR
+IN CONNECTION WITH THE WEIGHTS OR THEIR USE.
+
+A cleanly-licensed release trained only on permissively-licensed sources is planned.

README.md

@@ -1,5 +1,133 @@
 ---
 license: other
 license_name: pulse-research-preview
-license_link: LICENSE
+tags:
+  - decision-model
+  - text-classification
+  - zero-shot-classification
+  - on-device
+  - non-generative
+library_name: pytorch
 ---
+
+# Pulse Decide 150M
+
+**RouterML's on-device decision model.** 149.3M parameters (~570MB fp32). Non-generative:
+it *scores* typed decisions rather than generating text, which is why it runs on a laptop
+instead of a server.
+
+> Makes typed decisions — choice, boolean, multi-label, ordinal — over **arbitrary runtime
+> schemas**, offline, on-device.
+
+**Status: `v0.1` research preview.** This is an early research model, not a production release.
+
+![Pulse Decide — governed routing](pulse_beat.gif)
+
+*One customer message, routed twice. Without governed context Pulse isn't confident enough,
+so it escalates rather than guessing. With governed context it decides — 99.7% confident,
+on-device, in ~32ms.*
+
+## What it does
+
+Given a piece of text (the *state*) and a set of candidate labels supplied **at runtime**,
+Pulse scores each candidate and returns a decision plus a confidence. The label set does not
+need to be seen during training — that is the "arbitrary schema" property.
+
+## Key facts
+
+| | |
+|---|---|
+| Parameters | **149.3M** (0.149B) |
+| Encoder | `answerdotai/ModernBERT-base` |
+| Architecture | cross-encoder decision scorer (non-generative) |
+| Size on disk | ~570MB (fp32); ~150MB at int8 |
+| Runs on | CPU / Apple MPS / CUDA — **offline, no data leaves the device** |
+| Decision types | choice, boolean, multi-label, ordinal |
+
+## Performance
+
+Pulse is designed to run **behind a confidence gate** — answering what it is confident about
+and escalating the rest. That selective mode is how it is intended to be deployed, so it is
+what we report here. Measured on held-out, schema-disjoint evaluation sets (label sets never
+seen in training) at confidence threshold 0.5:
+
+| evaluation set | answering everything | confident subset only |
+|---|---:|---|
+| MASSIVE (intent, 60 labels) | 47.6% | **71.1%** at 47% coverage |
+| macro (3 held-out sets) | ~51% | **59.5%** at 63% coverage |
+
+Accuracy on *accepted* decisions is meaningfully higher than on forced answers — at the cost
+of escalating the remainder. This is the intended operating mode.
+
+> **A same-split benchmark evaluation against other published decision models is forthcoming.**
+> We make no claim of matching or exceeding any other model at this time.
+
+## Intended use
+
+- Research on small, non-generative decision models and governed routing
+- On-device / offline decisioning where data cannot leave the machine
+- As the local tier of an escalation system (cheap local decision → escalate when unsure)
+
+## Limitations
+
+- **Multi-label is weak** (~8% exact-set on FD multi-label heads). The model often identifies
+  the right labels but selects the wrong *number* of them. Active work.
+- **Confidence calibration drifts between training runs**; treat thresholds as needing
+  per-deployment calibration rather than as universal constants.
+- Not evaluated for fairness, safety, or high-stakes domains. **Do not use for consequential
+  decisions without your own evaluation.**
+- English only.
+- Early research preview — interfaces and weights may change.
+
+## Training data and licensing ⚠️
+
+Trained on a derived corpus assembled from **40 public text-classification datasets**
+(contamination-audited against all evaluation sets; near-duplicate filtered).
+
+**The source datasets carry mixed licenses**, including:
+- ~24 permissive (Apache-2.0, CC-BY, MIT-like)
+- several **research/non-commercial-only** datasets
+- several derived from **Twitter/X content**, whose Terms of Service restrict redistribution
+
+**We publish the model weights, not the training data.** Because the corpus mixes
+research-only and ToS-restricted sources, **users are responsible for verifying that their
+intended use complies with the underlying dataset licenses.** A cleanly-licensed
+(permissive-only) release is planned.
+
+## Usage
+
+```python
+import torch
+from pulse.model import PulseConfig, PulseUnified, load_tokenizer
+
+blob = torch.load("pulse_decide_150m.pt", map_location="cpu", weights_only=True)
+cfg = PulseConfig(**blob["cfg"])
+tok = load_tokenizer(cfg)
+model = PulseUnified(cfg); model.load_state_dict(blob["state_dict"]); model.eval()
+
+state = "My card was declined at the store this morning."
+choices = ["billing", "fraud", "technical", "account_access"]
+
+first = [f"{state} [SEP] Which category applies?"] * len(choices)
+enc = tok(first, choices, truncation=True, max_length=cfg.max_length,
+          padding=True, return_tensors="pt")
+with torch.no_grad():
+    scores = model.score_pairs(enc["input_ids"], enc["attention_mask"])
+print(choices[int(scores.argmax())])
+```
+
+## Model family
+
+Pulse Decide is the decision model inside **RouterML**, which adds the governance layer —
+eligibility, confidence gating, escalation and verification — around it.
+
+## Citation
+
+```bibtex
+@software{pulse_decide_150m,
+  title  = {Pulse Decide 150M: an on-device non-generative decision model},
+  author = {RouterML},
+  year   = {2026},
+  note   = {v0.1 research preview}
+}
+```

config.json

@@ -0,0 +1,16 @@
+{
+  "model_name": "Pulse Decide 150M",
+  "developer": "RouterML",
+  "architecture": "cross-encoder decision scorer (non-generative)",
+  "encoder": "answerdotai/ModernBERT-base",
+  "parameters": 149300000,
+  "version": "v0.1-research-preview",
+  "decision_types": [
+    "choice",
+    "boolean",
+    "multi_label",
+    "ordinal"
+  ],
+  "checkpoint_sha256": "424c9e2bf6acda3ef792cae8d8b7cfe70829633893b5d0af250e8259b3832f95",
+  "checkpoint_file": "pulse_decide_150m.pt"
+}

pulse_decide_150m.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:424c9e2bf6acda3ef792cae8d8b7cfe70829633893b5d0af250e8259b3832f95
+size 597307292