SRT-Adapter v8a — Live Demo

Interactive demo for the Semiotic-Reflexive Transformer Adapter (v8a generation) bolted onto a frozen Qwen/Qwen2.5-7B.

Paste a passage. Get back four readouts:

1. Per-token $\hat{r}$ heatmap — continuous reflexivity / contestedness
2. Per-token $P(\text{supercritical})$ — binary regime probability
3. Layer-wise divergence norms — where meaning forks across discourse communities (layers 7 / 14 / 21)
4. Community vector summary — 64-D latent community embedding (norm + top dims)

This Space targets ZeroGPU (free Nvidia A100 minutes via the community grant). First request after a cold start downloads Qwen2.5-7B (~15 GB) into the Space cache, which can take a few minutes; subsequent requests return in under a second.

For the paper, weights, and reproducible benchmarks, see the model card.