JADE

Checkpoints	Inputs
theunnecessarythings/JADE	text

JADE is a LoRA adapter on Qwen/Qwen3.8-27B. Flide reads the adapter the release serves with vLLM, in vllm-adapter/: rank-8 LoRAs on the projections and a rank-256 LoRA on the output head, all at the scaling its configuration gives (alpha over r). Flide runs them beside the base weights and reads the output head's rows of the answer codes with their LoRA, with no extra package.

Flide asks it with the release's own prompt. Each question is a chat prompt of its own. The system turn asks for the code of the selected option and tells the model to treat the state as data. The user turn holds "State:" with the state, "Question:" with the question, and "Options:" with one line per option, and asks for the letter code of the best option. A state that is not a string is written as JSON. A choice lists each option's name and, when it has one, its description. A noul lists "No / false" and then "Yes / true" unless the request describes the two outcomes. A question without instructions is asked as "Choose the best matching option.".

The options are labelled with the codes of decision_config.json, and the answer is the softmax over the codes of the question's options at the first token of the reply, at the temperature that decision_config.json sets. The state comes first in every prompt, so Flide computes it once and forks it into the questions. Thinking is off.

Speed

On one H200, one request at a time, Flide answers 1.7 to 9.2 times faster than JADE's own JadeEngine on vLLM 0.30. Each workload is 32 requests, each run once on both sides before timing, and the table gives the median latency of a request. MultiRC requests ask about 49 questions each. The release keeps a prefix cache across requests, so it is emptied after that first run and the timed requests are read afresh on both sides.

Workload	Flide	Release	Speedup
Decision Index	61.6 ms	123.3 ms	2.0×
RACE	44.1 ms	102.5 ms	2.3×
MultiRC	368.5 ms	3379.9 ms	9.2×
QuALITY long	367.6 ms	642.2 ms	1.7×

Start a server

flide serve --model theunnecessarythings/JADE

Send a request

curl http://localhost:8000/v1/systemone -H "Content-Type: application/json" -d '{
  "state": "Order #1182 arrived with a cracked screen. The customer asks for a refund and attached two photos.",
  "questions": {
    "action": {"type": "choice", "instructions": "What should support do next?",
               "criteria": {"refund": "refund the order", "replace": "send a replacement", "escalate": "ask a human"}},
    "urgent": {"type": "noul", "instructions": "Does this need a reply within the hour?"}
  }
}'

Limits and options

A request takes choice and noul questions, and a score is refused. A choice takes up to 255 options. The longest prompt is 8191 tokens; a longer prompt is refused. A request with images is refused.

Option	What it does
--format template --format-arg spec=theunnecessarythings__jade	reads a copy of the checkpoint kept in a directory Flide does not recognize