README.md

---
library_name: transformers
tags:
- qualitative-coding
- cpu
- text2text-generation
extra_gated_prompt: >-
  These checkpoints were trained on a qualitative-coding corpus whose
  redistribution licence and participant-consent basis are not documented in
  the retained research artifact (see "Data provenance" below). Access is
  granted for non-commercial research use and reproduction of the accompanying
  paper only. You are responsible for confirming that your own use is lawful in
  your jurisdiction. Do not redistribute the weights or attempt to reconstruct
  the underlying source corpus.
extra_gated_fields:
  Name: text
  Affiliation: text
  Intended use: text
  I will use these weights for non-commercial research only: checkbox
  I will not redistribute the weights or reconstruct the source corpus: checkbox
extra_gated_button_content: Request access
---


# CPU Open-Coding Models

This repository archives five CPU-capable English
open-coding checkpoints from the matched qualitative-coding study.

| Subfolder | Architecture | Intended status |
|---|---|---|
| `tiny-coder-v1` | T5-Efficient-Tiny, 15.6M | early clean coder |
| `tiny-hybrid-v2` | T5-Efficient-Tiny, 15.6M | hybrid tiny candidate |
| `mini-coder-v1` | T5-Efficient-Mini | intermediate-size baseline |
| `flan-small-coder` | FLAN-T5-Small | clean FLAN baseline |
| `flan-small-hybrid-v2` | FLAN-T5-Small | interactive website candidate |

Load a checkpoint with a subfolder:

```python
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer

repo = "Likich/cpu-open-coding-models"
subfolder = "tiny-hybrid-v2"
tokenizer = AutoTokenizer.from_pretrained(repo, subfolder=subfolder)
model = AutoModelForSeq2SeqLM.from_pretrained(repo, subfolder=subfolder)
```

These models generate short first-pass code suggestions. Outputs must remain
editable and reviewable; the models do not perform complete qualitative
analysis and should not be treated as substitutes for researchers.

The expert evaluation found a human-retention benefit but also a substantial
quality gap from task-trained Qwen2.5-7B. Source-data permissions are still
under review. Keeping this repository private does not authorize onward
distribution of its weights or training examples.

## Which checkpoint to use

`tiny-hybrid-v2` (T5-Efficient-Tiny, 15.6M) is the CPU model reported in the
paper: 61.7 MiB, a median 0.021 s per passage and 444 MiB peak process RSS on a
2021 M1 Max. `flan-small-hybrid-v2` is the larger FLAN candidate.

## Data provenance and release status

These weights derive from a 999-pair English open-coding benchmark: 600 pairs
from social-science work across three university faculties (interviews and
reviews, consensus-coded by three to five coders) and 399 SemEval-2014 Task 4
review excerpts, plus 1,990 machine-coded ICLR peer-review excerpts.

The supplied artifact records only the passage and its label. It does **not**
record the original language, coder identities, adjudication trace, consent
basis, or redistribution licence. Public availability of source text does not
by itself establish permission to redistribute a compiled corpus or weights
trained on it. Access is therefore gated, and the source passages are **not**
released.

## Memorisation

These are small sequence-to-sequence models fitted to a small label set, and
they reproduce training annotations verbatim at a substantial rate. At the
matched epoch-8 checkpoint, **28.9% of human-code (HR) outputs across 1,000
evaluated items are exact strings from the human training labels**, drawing on
28 distinct labels. Treat generated codes as potentially reproducing the
original coders' annotations rather than as novel interpretations. Long source
passages are not recoverable from a model of this size; the annotation set is
partially recoverable, which is why access is gated.

## Intended use and limits

First-pass, editable code suggestions for one pre-segmented English passage,
returning one code of at most six words. These models do not segment
transcripts, assign multiple codes, build codebooks, or construct themes, and
are not a substitute for a researcher. In a blinded five-expert evaluation,
47.5% of Tiny-HR suggestions were rated usable or better, against 85.0% for a
task-adapted Qwen2.5-7B.

## Citation

Accompanying paper: *AI-Assisted Qualitative Coding on a CPU* (under review).
Code, prompts, analysis scripts and hashes accompany the submission.