# Show HN: An open-weight LLM whose answer/stop decision can be flipped internally

> Source: <https://github.com/theonlypal/PCCG-Qwen3-4B-continuation-control>
> Published: 2026-09-09 22:32:34+00:00

**40/40 GO→EOS**

**40/40 EOS→GO**

**640/640 controls unchanged**

**Prompt, weights, and generated reasoning prefix held fixed during intervention.**

Rayan Pal

[getswiftapi.com](https://getswiftapi.com)

Independent Researcher

Thinking-enabled [PCCG-Qwen3-4B](https://huggingface.co/sharthokrayanpal/PCCG-Qwen3-4B) with an equality-conditioned continuation policy. One fixed 2,560-dimensional activation direction at block 29, applied after the model generated `</think>`. Intervention strength: 2.0. `GO` is followed by native EOS.

| Confirmation measurement | Result | 
|---|---|
| Correct generated comparison trace and answer | 80/80 | 
| Unmodified replay | 80/80 | 
| GO→EOS | 40/40 | 
| EOS→GO | 40/40 | 
| Controls unchanged | 640/640 | 

Anthropic's Jacobian Lens provides internal vocabulary readouts. Additive activation interventions measure causal effects. [Method](/theonlypal/PCCG-Qwen3-4B-continuation-control/blob/main/METHOD.md).

Clone with Git LFS installed. The fitted lens is stored in Git LFS.

```
git lfs pull
python verify.py
```

Checks package hashes and recounts the confirmation records. Runs on CPU using the Python standard library.

[Commands and model requirements](/theonlypal/PCCG-Qwen3-4B-continuation-control/blob/main/REPRODUCE.md).

- `results/` : complete study records, logits, activations, fitted lens, and fixed direction.
- `source/` : executed study code and Anthropic's Jacobian Lens implementation.
- `protocol/` : prompts, case sets, model qualification record, and reproduction settings.
- `SHA256SUMS` : package file hashes.
- `LICENSE` : Apache 2.0. The vendored implementation retains its license in`source/vendor/LICENSE` .
