chadstthomas/Paraphase-4B-Adapter-Pack

Paraphase-4B Adapter Pack — rewrite + extraction LoRAs

One raw Qwen3.5-4B base, task LoRAs, and one measured tool guard. This is the multi-task companion to Paraphase-4B, which is the merged humanizer release.

Use the raw base here for adapters. The paraphase-4b-Q8_0.gguf in the main repo is the merged rewrite model; it is not the adapter base. The raw qwen3.5-4b-Q8_0.gguf in this pack is the base the 4B LoRAs were trained on. The 2B extraction LoRA applies to raw openbmb/MiniCPM5-2B (see below).

Quick start

# extraction (4B)
llama-server -m qwen3.5-4b-Q8_0.gguf \
  --lora gguf/paraphase-4b-extract-lora.gguf -c 2048

# rewrite (4B)
llama-server -m qwen3.5-4b-Q8_0.gguf \
  --lora gguf/paraphase-4b-rewrite-lora.gguf -c 2048

# extraction (2B, on-device tier; official MiniCPM5-2B GGUF)
llama-server -m MiniCPM5-2B-Q8_0.gguf \
  --lora gguf/paraphase-2b-extract-lora.gguf -c 2048

# tool calls: base only; apply toolguard.py to the returned calls
llama-server -m qwen3.5-4b-Q8_0.gguf -c 2048

Send "chat_template_kwargs": {"enable_thinking": false} with Qwen3.5 requests; otherwise the template starts with a thinking preamble.

What is measured

Extraction adapter (the win)

Frozen test: 50 real AESLC validation emails, disjoint from training. Labels are deterministic regex spans; this measures agreement with that labeler, not human-judged extraction. Labeler v2 (2026-10-23): amount spans now include magnitude words ($12 million, 1.2 million dollars, $12k), fixing the one flaw in v1 — same 50 emails, re-labeled; the shipped adapters are trained on v2 labels and measured on extract_smoke.v2.

metric	raw Q8 base	+ extract LoRA
JSON valid	1.00	1.00
exact match	0.12	0.40
micro F1	0.522	0.674
date F1	0.407	0.868
amount F1	0.730	0.722
action-item F1	0.522	0.658
empty-row accuracy	0.40	0.80
non-empty recall	0.756	0.889

The adapter raises exact match from 6/50 to 20/50 and doubles date F1. Under v2 labels amount F1 becomes meaningful for the first time: the raw base is already strong on amounts (0.730) and the adapter holds it (0.722) while gaining everywhere else. Only 4/50 test emails contain amounts, so the amount row still carries less weight than dates/actions. The v1 numbers (exact 0.42, micro 0.670 on extract_smoke.v1) remain on record in the project docs.

2B extraction adapter (on-device tier)

Same recipe on raw openbmb/MiniCPM5-2B (150 steps, ~2 minutes of training, peak 6.49 GiB), trained and measured on the same v2 labels — at one-third the download (2B Q8 1.4 GB + 50 MB LoRA):

metric	2B base	2B + extract LoRA	4B + extract LoRA (reference)
JSON valid	0.96	1.00	1.00
exact match	0.14	0.34	0.40
micro F1	0.277	0.631	0.674
date F1	0.313	0.806	0.868
amount F1	0.438	0.766	0.722
action F1	0.272	0.614	0.658
non-empty recall	0.378	0.867	0.889

Base GGUF: openbmb/MiniCPM5-2B-GGUF (MiniCPM5-2B-Q8_0.gguf). MiniCPM5 has no thinking preamble, so no enable_thinking flag is needed.

Tool path (base + guard)

Frozen 20-item tool-call smoke:

candidate	accuracy
raw Q8 base	0.70
raw base + toolguard.py	0.75
tool LoRA (withdrawn)	0.65

The tool LoRA fixed one wrong-tool case but made two web-search queries more verbose, so it is not included. The guard is a narrow deterministic fix: when a request starts with Email/Send, contains an explicit address, and the model chose get_contact_email, it converts the call to send_email with the parsed to/subject/body. It changes exactly one frozen item and never touches web search or no-tool cases.

Rewrite adapter

The rewrite LoRA is the adapter merged into Paraphase-4B. The main model card has the voice50 and Pangram evidence: 50/50 valid, fact fidelity 0.970, copy reuse 0.260 with the retry policy, tell rate 0.08.

Files

file	size	notes
qwen3.5-4b-Q8_0.gguf	4.3 GB	raw Qwen3.5-4B base; Q8_0
gguf/paraphase-4b-extract-lora.gguf	42.5 MB	extraction adapter, llama.cpp
gguf/paraphase-4b-rewrite-lora.gguf	42.5 MB	rewrite adapter, llama.cpp
gguf/paraphase-2b-extract-lora.gguf	50.3 MB	2B extraction adapter (base: openbmb/MiniCPM5-2B)
peft/rewrite/	85 MB	PEFT adapter for transformers
peft/extract/	85 MB	PEFT adapter for transformers
peft/extract-2b/	101 MB	PEFT 2B extraction adapter
toolguard.py	3 KB	narrow send/contact guard
task-adapter-manifest.json	2 KB	hashes, sizes, measured endpoints
LICENSE	Apache-2.0	inherited from Qwen3.5-4B

Extraction prompt

System:

You extract structured data from email. Return only valid JSON.

User:

Extract the dates and times, money amounts, and action items from the email.
Return only a JSON object with keys "dates", "amounts", and "action_items".
Dates and amounts must be strings copied exactly from the email.
Each action item is the sentence that states a request or task.
Use [] when a field has none.

EMAIL:
<subject and body>

The adapter returns JSON such as:

{"dates": ["Thursday", "3pm", "Friday"], "amounts": ["1,240 dollars"], "action_items": ["Please send the signed copy by Friday."]}

Tool guard

from toolguard import guard_tool_call

calls, changed = guard_tool_call(user_text, model_tool_calls)

The guard only handles the explicit-address send/contact confusion. It is not a general tool router and it does not repair arbitrary argument errors.

PEFT / transformers

from transformers import AutoModelForCausalLM, AutoTokenizer
from peft import PeftModel

base = "Qwen/Qwen3.5-4B"
tok = AutoTokenizer.from_pretrained(base)
model = AutoModelForCausalLM.from_pretrained(base, device_map="auto")
model = PeftModel.from_pretrained(model, "peft/extract")

Load the base in bf16 or 4-bit; the adapter is r=16, alpha=32.

Reproduce

# raw base
python convert_hf_to_gguf.py <Qwen3.5-4B snapshot> \
  --outfile qwen3.5-4b-Q8_0.gguf --outtype q8_0

# extraction: 427 train / 48 val AESLC, 150 steps
PYTHONPATH=. python scripts/train_task_adapter.py --task extract \
  --data data/processed/extract_v1/train.jsonl \
  --out data/checkpoints/extract-adapter --steps 150

# extraction eval
PYTHONPATH=. python scripts/eval_extract.py \
  --gguf qwen3.5-4b-Q8_0.gguf \
  --lora gguf/paraphase-4b-extract-lora.gguf --tag extract-adapter

# tool guard (stored base outputs, no model call)
PYTHONPATH=. python scripts/eval_tool_guard.py

The full route and data provenance are summarized in task-adapter-manifest.json and the main Paraphase-4B card.

Limitations

• Extraction labels are regex-derived, not human annotations. Amount F1 is low-confidence; dates and action items carry the result.
• Action-item labels are heuristic sentences; the adapter can beat or lose to the labeler for reasons unrelated to extraction quality.
• The tool guard fixes one demonstrated case; the strict tool instrument is mostly a query-phrasing test.
• One LoRA per llama-server process; adapter hot-swapping is not implemented.
• No detector was used for these adapters. Pangram evidence belongs to the rewrite model, not the extraction task.

License

Apache-2.0, inherited from Qwen3.5-4B. The base model is Qwen/Qwen3.5-4B.

Links

• Paraphase collection — the family and the premise
• Paraphase-4B (merged humanizer)
• Live rewriter
• Interactive comparison
• Paraphase-2B