File size: 16,114 Bytes
08f3aef 551f3ba 08f3aef 551f3ba ce07254 551f3ba e71d16d 551f3ba e71d16d 08f3aef 551f3ba e71d16d 08f3aef e71d16d 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba 08f3aef 551f3ba e71d16d ce07254 551f3ba e71d16d ce07254 26c7a37 ce07254 7e8389e ce07254 e71d16d ce07254 e71d16d ce07254 e71d16d 26c7a37 e71d16d 26c7a37 e71d16d 08f3aef e71d16d 08f3aef e815622 ce07254 e71d16d 08f3aef 8af51bb e71d16d 4d5d03d ce07254 4d5d03d 08f3aef ce07254 | 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 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 | ---
language:
- en
- zh
license: apache-2.0
library_name: transformers
pipeline_tag: text-classification
tags:
- decision-making
- typed-decisions
- text-classification
- small-model
- bilingual
- calibration
- option-order-invariance
- systemone
- zero-output-tokens
- on-premise
base_model: jhu-clsp/mmBERT-small
datasets:
- LocalLLaMA/typed-decisions
thumbnail: figures/C1_typed_acc_comparison.png
widget:
- text: "State: The system processed 3 invoices without errors.\nQuestion: Which action should be taken next?\nOptions: A) Approve batch B) Retry the batch C) Flag for human review"
example_title: "choice — pick one of three"
output:
- label: "A) Approve batch"
score: 0.62
- label: "B) Retry the batch"
score: 0.24
- label: "C) Flag for human review"
score: 0.14
- text: "State: The user asks the agent to delete the production database and purge all backups.\nQuestion: Should this action be allowed?\nAnswer: yes or no."
example_title: "noul — allow/deny gate"
output:
- label: "no"
score: 0.93
- label: "yes"
score: 0.07
- text: "State: The candidate's resume shows 5 years of backend engineering experience.\nQuestion: Rate how well the candidate matches criterion 'experience'.\nAnswer: 2 (worst) to 10 (best)."
example_title: "score — 2..10 rating"
output:
- label: "6"
score: 0.72
---
<p align="center">
<img alt="License" src="https://img.shields.io/badge/License-Apache%202.0-blue?style=flat-square">
<img alt="Params" src="https://img.shields.io/badge/Params-144.3M-orange?style=flat-square">
<img alt="Typed ACC" src="https://img.shields.io/badge/Typed%20ACC-en%200\.797%20(self-host%20eval)-green?style=flat-square">
<img alt="Latency" src="https://img.shields.io/badge/Latency-21.0ms%20GPU%20fp16%20RTX%205090-9cf?style=flat-square">
<a href="https://github.com/Phocinae/Phocinae-Largha-150M-v1"><img alt="GitHub Stars" src="https://img.shields.io/github/stars/Phocinae/Phocinae-Largha-150M-v1?style=flat-square&logo=github"></a>
<img alt="HF Downloads" src="https://img.shields.io/badge/dynamic/json?url=https%3A%2F%2Fhuggingface.co%2Fapi%2Fmodels%2FPhocinae%2FPhocinae-Largha-150M-v1&query=%24.downloads&label=HF%20Downloads&color=orange&style=flat-square">
<a href="https://phocinae.github.io/Phocinae-Largha-150M-v1/"><img alt="Site" src="https://img.shields.io/badge/Site-live-brightgreen?style=flat-square"></a>
</p>
<div align="center">
<img src="figures/logo_phocinae.png" width="180" alt="斑海豹 Phocinae logo">
<h1>斑海豹 · Phocinae-Largha-150M-v1</h1>
<p><strong>小海豹,大决断。</strong> / <em>Tiny model. Big decisions.</em></p>
<p><em>一斑见全豹,一点定全局。</em> / <em>Spotted seal. Spot-on calls.</em></p>
</div>
斑海豹 **Largha**, the spotted seal: a **144.3M decision model** (en-first; zh evaluated on machine-translated cases) (150M-class) for structured decisions — one forward pass per decision, on your own hardware. Not a chat model: it takes a `state` plus a list of typed questions (`noul` yes/no · `choice` pick-one · `score` 2–10) and returns calibrated answers with confidence, robust to option reordering. Downloads: [Hugging Face](https://huggingface.co/Phocinae/Phocinae-Largha-150M-v1) · [GitHub](https://github.com/Phocinae/Phocinae-Largha-150M-v1) · [ModelScope 魔搭](https://modelscope.cn/models/PerryLink/Phocinae-Largha-150M-v1) · 中文: [FAQ 中文版](docs/faq.zh.md) · [场景演示画廊](docs/gallery/README_cn.md).
## TL;DR
| TL;DR | value |
|---|---|
| parameters | **144.3M** (mmBERT-small base: hidden 384 × 22 layers × 6 heads, 256k-vocab tokenizer, RoPE + sliding-window + full attention; base context 8192; decision sequence ≤512 tokens with a 192-token head attention window) |
| typed-decisions en (400 cases / 2000 decisions) | **0.906** (specialist: fitted on this dataset's train split) — Laya 0.766 (self-measured, native interface) · JEV 0.727 (generalist, zero-shot) · meraGPT 0.768 (generalist, zero-shot) |
\| typed-decisions zh (machine-translated cases; training mix includes machine-translated Chinese ≈2,400 rows and native Chinese ≈1,400 rows) \| **0.848** \|
| option-order flip robustness (lower is better) | CPU fp32: flip150 **0.0200** · flip400 **0.0217** · random-mean **0.0144** · any **0.0283** (GPU fp16 0.0200/0.0217, note only) |
| inference latency | GPU fp16 p50 **21.0 ms** (RTX 5090) · CPU single-thread p50 **1.64 s per case** (1 case = 1 state + 5 questions, single forward pass) · CPU 8-thread batch **8–20 decisions/s** |
| JevBench public-231 | **0.5455** (126/231) — below the 58.4% gate, disclosed honestly; tool_selection 12/12 (n=12) |
| escalate routing (E1 gate, τ=0.6) | **+0.0876 kept-subset acc** (0.906→0.9936 (kept subset)) while **−55.0% LLM calls (79.6% at τ=0.5)** (45.0% official-set escalate; τ swept on the eval set — re-scan per domain) |
| calibration | shipped column ECE **0.2519** (en) / **0.1941** (zh); with the bundled calibration column: **0.0168** / **0.0152**; calibration temperatures 0.8660205/0.8081192/0.6624661 (applied at inference) |
Full numbers, methodology, and evidence: **[BENCHMARKS.md](./BENCHMARKS.md)**.
## See it in action
<div align="center">
<img src="figures/C1_typed_acc_comparison.png" width="600" alt="typed accuracy comparison"/>
<p><em>图 1 · typed-decisions accuracy: 0.906 (en) / 0.848 (zh) — Laya 0.766 (self-measured, native) · JEV 0.727 · meraGPT 0.768</em></p>
<img src="figures/C7_routing_savings.png" width="600" alt="routing savings"/>
<p><em>图 2 · escalate routing (τ=0.6): accuracy 0.906→0.9936 (kept subset), LLM calls −55.0% (79.6% at τ=0.5)</em></p>
<img src="docs/gallery/G25_race_local_vs_api.gif" width="600" alt="local vs API race"/>
<p><em>图 3 · 同一决策:本地 21.0 ms (RTX 5090) vs API 往返 ≈1.5 s(我方实测 n=40;第三方实测 Jev API 单决策 238–301 ms)</em></p>
<img src="docs/gallery/S01_rmrf_gate.gif" width="600" alt="command gate"/>
<p><em>图 4 · 命令审批门:rm -rf 在 21.0 ms (RTX 5090) 内拦截,p(deny)=0.96(合成演示数据)</em></p>
</div>
**📽️ 全部 27 个场景演示 → [docs/gallery/](./docs/gallery/)**(审批安全 · 路由省费 · 实时分级 · 办公文档 · 流程工程 · 对照与可靠性;中文版 [gallery/README_cn.md](./docs/gallery/README_cn.md))
## Why Phocinae: slash agent costs
**55.0% fewer LLM calls, 21.0 ms per decision on an RTX 5090.** Phocinae-Largha-150M (144.3M params) routes the repetitive decisions inside agent sessions — command approvals, tool selection (12/12 on JevBench tool_selection, k≤10), step checks, output screening — to a local single-forward-pass engine (8192-token base context, 512-token default head) instead of a 500–4,000-token API call. The τ=0.6 confidence gate cuts LLM traffic 55.0% with kept-subset accuracy 0.906 → 0.9936 and full routed accuracy 0.8135; deterministic inference means auditable, replayable decisions with zero data leaving your machine — at 21.0 ms GPU p50 (RTX 5090) vs 1.5 s+ API round-trips. Full cost model: [docs/cost-savings.md](./docs/cost-savings.md).
<div align="center">
<img src="figures/C2_latency_comparison.png" width="600" alt="latency comparison"/>
<p><em>图 5 · per-decision latency: GPU fp16 21.0 ms (RTX 5090) · CPU single-thread 1.64 s/case · CPU 8-thread batch 8–20 decisions/s</em></p>
<img src="figures/C8_params_vs_latency.png" width="600" alt="params vs latency"/>
<p><em>图 6 · 144.3M 参数、288.6 MB 权重、1.6 GB 推理峰值显存 — 一张普通显卡即可本地运行</em></p>
<img src="figures/C6_bilingual.png" width="600" alt="bilingual"/>
<p><em>图 7 · 双语决策:同一判定中英并排 en 0.906 / zh 0.848</em></p>
</div>
## Quick start — phocinae-server
> `phocinae-server` ships the **runtime**, not the weights — fetch the weights first (step 1).
The official runtime is **[phocinae-server](https://github.com/Phocinae/phocinae-server)**: a local FastAPI service (127.0.0.1 only) that loads these weights with a pure PyTorch forward (no extra runtime needed). Full spec: [docs/protocol.md](./docs/protocol.md). Hardware tiers: [docs/deployment.md](./docs/deployment.md).
```bash
# 1) get the weights (either CLI works; `hf` is the newer one)
hf download Phocinae/Phocinae-Largha-150M-v1 --local-dir ./largha
# older huggingface_hub releases:
# huggingface-cli download Phocinae/Phocinae-Largha-150M-v1 --local-dir ./largha
# 2) install the runtime and point it at that directory
pip install phocinae-server
PHOC_MODEL_DIR=./largha python -m phocinae.main # http://127.0.0.1:8155
```
The directory produced by that download is exactly the layout `PHOC_MODEL_DIR`
expects (`model.safetensors`, `encoder/`, `tokenizer/`, `rl_agent_config.json`).
From Python (the same engine the server uses):
```python
from phocinae.engine import Engine
eng = Engine("./largha", device="cpu") # device="auto" selects CUDA when present
answers, confidence, action, usage = eng.run(
"The agent restarted nginx after checking the logs and the health endpoint is green.",
[{"id": "ok", "type": "noul"},
{"id": "act", "type": "choice", "options": ["allow", "ask", "deny"]},
{"id": "risk", "type": "score"}],
)
# answers -> {'ok': False, 'act': 0, 'risk': 3}
# confidence -> {'ok': 0.668, 'act': 0.409, 'risk': 0.1633}
# usage -> {'input_tokens': 146, 'output_tokens': 0}
# pass with_scores=True to get a 5th value (per-option scores)
```
One decision over HTTP:
```bash
curl -s http://127.0.0.1:8155/v1/systemone -H 'Content-Type: application/json' -d '{
"model": "Phocinae-Largha-150M-v1",
"state": "The agent restarted nginx after checking the logs and the health endpoint is green.",
"questions": [
{"id": "ok", "type": "noul", "threshold": 0.65},
{"id": "act", "type": "choice", "options": ["allow", "ask", "deny"]},
{"id": "risk", "type": "score"}
]
}'
```
```json
{
"model": "Phocinae-Largha-150M-v1",
"answers": {"ok": false, "act": 0, "risk": 3},
"usage": {"input_tokens": 146, "output_tokens": 0},
"answer_confidence": {"ok": 0.668, "act": 0.409, "risk": 0.1633},
"action": {"act": {"act_probability": 0.1581}},
"routing": {"model": "Phocinae-Largha-150M-v1", "device": "cpu", "perm": "none", "backend": "phocinae-pure-torch"}
}
```
> **Known limitation in tool routing.** `Router.route_tool()` confuses semantically
> close tool names. Reproducible case: state *"The user wants to find the latest news
> about the product launch."* with the six tools `web_search / read_file / run_command /
> list_files / fetch_url / ask_user` returns **`fetch_url`**, not `web_search`, at
> confidence ≈0.33. Treat low-confidence tool picks as escalate-worthy rather than final.
`answers` values: `noul` = bool · `choice` = 0-based option index · `score` = integer 2–10. Errors: **422** (unknown model/type, >64 questions, choice without options, score with options, threshold outside [0,1]) · **413** (>2 MiB body) · **401** (bearer token). Extension keys `answer_confidence` / `action` / `routing` can be disabled with `PHOC_EXTENSIONS=0`.
## Documentation
- **[BENCHMARKS.md](./BENCHMARKS.md)** — full comparison tables, charts, methodology, honest disclosures
- **[MODEL_CARD.md](./MODEL_CARD.md)** — detailed model card: architecture, training, evaluation, bias & limitations
- **[docs/deployment.md](./docs/deployment.md)** — serve, guard, MCP integration; hardware tiers; security notes
- **[docs/protocol.md](./docs/protocol.md)** — the `/v1/systemone` decision protocol
- **[docs/reproduce.md](./docs/reproduce.md)** — evaluation protocols, published numbers, evidence paths
- **[docs/cost-savings.md](./docs/cost-savings.md)** — LLM-cost model for the escalate gate
- **[docs/gallery/](./docs/gallery/)** — 27 application scenarios with animated demos (中文: [gallery/README_cn.md](./docs/gallery/README_cn.md))
- **[calib/](./calib/)** — optional post-hoc calibration column (power transform; ECE 0.2519/0.1941 → 0.0168/0.0152)
- **[docs/faq.md](./docs/faq.md)** — common questions (中文: [faq.zh.md](./docs/faq.zh.md)) · **[docs/technical-report.md](./docs/technical-report.md)** — short technical report
## What it is / what it is not
- **For**: structured decisions — approval gates, tool routing, escalation, document triage, step checks, output screening; anywhere you want a fast, cheap, local decision layer instead of a large model.
- **Not for**: open-ended chat / generation, long-document reasoning, or world-knowledge QA (MMLU-style probes are below par; see MODEL_CARD). Not a safety oracle: use as a first-line gate with escalation, never as the sole guard.
## Honest disclosures
- **JevBench public-231: 0.5455 (126/231) vs a 58.4% acceptance gate — not passed.** We publish the number as measured, and we never train on the eval rows.
- zh results are on machine-translated cases; the training mix includes machine-translated Chinese (≈2,400 rows) and native Chinese (≈1,400 rows) — treat zh as an in-mix (fitted) evaluation, not zero-shot cross-lingual transfer.
- Flip numbers are measured per option-reorder protocol (lower is better): CPU fp32 0.0200/0.0217 (flip150/400 reversed) · random-mean 0.0144 · any 0.0283. GPU fp16 0.0200/0.0217 and 1k-row 0.0181/0.0150/0.0331 are note-only values from different protocols.
- Calibration: the shipped column has ECE **0.2519** (en). Calibration temperatures (0.8660205/0.8081192/0.6624661) are stored in the model repo config and applied at inference by phocinae-server. A recommended recalibration column is bundled under [`calib/`](./calib/).
## Weights & license
- `model.safetensors` sha256 `b6472511eea30729985374f43968cf7f4b16bbe0827de6c6ca07cf92afbb778a` (fp16 storage, 144.3M params); trained from `jhu-clsp/mmBERT-small` on `LocalLLaMA/typed-decisions` (train split + flip-augmented reorderings), recipe in `rl_agent_config.json`.
- **Apache-2.0** (see LICENSE); base encoder `jhu-clsp/mmBERT-small` is MIT (see NOTICE). · Repo layout: `model.safetensors` · `encoder/` · `tokenizer/` · `rl_agent_config.json` · `checkpoint_meta.json` · `figures/` · `docs/`.
## Credits
- [typed-decisions](https://huggingface.co/datasets/LocalLLaMA/typed-decisions) (Apache-2.0, LocalLLaMA HF org) — protocol & test data
- [mmBERT-small](https://huggingface.co/jhu-clsp/mmBERT-small) (JHU CLSP) — base encoder
- [JevBench](https://github.com/fstandhartinger/JevBench) — held-out protocol used for disclosure
## Version & machine-readable sources
- **This release**: `v1.1` — pinned tag on the model repository (`v1.0.0` marks the previous release), for reproducible citation.
Resolve its commit with `git ls-remote --tags https://huggingface.co/Phocinae/Phocinae-Largha-150M-v1`.
(No commit hash is written into this file on purpose: it would go stale on the next commit.)
- **Machine-readable facts** (same numbers as this card, for AI systems and retrieval pipelines): [llms.txt](./llms.txt)
- **Citation metadata**: [CITATION.cff](./CITATION.cff)
- **Landing page** (mirrors this card, with structured data): https://phocinae.github.io/Phocinae-Largha-150M-v1/
## Citation
Machine-readable: [CITATION.cff](./CITATION.cff). BibTeX:
```bibtex
@misc{phocinae2026largha,
title = {Phocinae-Largha-150M-v1: A 144M-parameter bilingual typed decision model},
author = {Phocinae Project},
year = {2026},
url = {https://huggingface.co/Phocinae/Phocinae-Largha-150M-v1},
license = {Apache-2.0}
}
```
## Revision history
- **2026-10-09 — v1.1 refresh.** Weights upgraded (each metric in this card re-measured on the new weights; previous release sha256 `db79d5ee2f16597f34e564f5a4363bddb5b5bbd9827c01819725dabcc7802697`). Figures/gallery re-rendered; optional calibration column added under [`calib/`](./calib/).
- **2026-10-09 — docs.** Expanded quick start (weight download, Python & HTTP examples, tool-routing caveat); added landing page & machine-readable sources section.
|