# Researchers compose frozen RWKV and Pythia models via shared latent adapter

Digest AI · Research · published 2026-10-01T04:00:00Z

Canonical: https://digestai.news/story/researchers-compose-frozen-rwkv-and-pythia-models-via-shared-latent-ad

## Summary

A new arXiv paper introduces NinaXander, a method that connects layers of two frozen language models from different architecture families using a single trained adapter. The study combines RWKV-4-Raven-7B, a recurrent model, with Tulu-Pythia-6.9b, a Transformer-based model. Once the adapter is trained, multiple composed models can be created by connecting at different layer boundaries without retraining.

The best configuration uses the first 5 layers of Pythia and the remaining 27 layers of RWKV. This setup reduces the Transformer key-value cache by 84.4% while maintaining accuracy comparable to RWKV alone on multiple-choice tasks. However, no composed model matches Pythia's accuracy, and language modeling performance drops sharply on WikiText, an out-of-domain corpus. The authors note that representation alignment was only achieved under favorable conditions — shared tokenizer, depth, and hidden width — and does not prove a general shared semantic space.

## Key points

- NinaXander connects frozen RWKV and Pythia models via a shared latent adapter
- Best composition cuts KV cache 84.4% with RWKV-level multiple-choice accuracy
- No composed model matches Pythia accuracy; WikiText perplexity degrades sharply

## Why it matters

Shows post-hoc model composition across architectures is feasible for inference efficiency, but reveals limits in preserving capabilities and generalization.

## Sources

1. [NinaXander: Feasibility and Limits of Composing Frozen Language Models Across Architecture Families via a Shared Latent Space](https://arxiv.org/abs/2609.38261) (arXiv cs.CL, 2026-10-01, primary source)

## Cite

Digest AI, "Researchers compose frozen RWKV and Pythia models via shared latent adapter", 1 October 2026, https://digestai.news/story/researchers-compose-frozen-rwkv-and-pythia-models-via-shared-latent-ad

---

Written by Digest AI's editorial model from the linked sources; the sources are the record. Headlines, digests and key points are written by Digest AI and may be quoted with a link to the story page. Linked articles belong to their publishers. Terms: https://digestai.news/terms#reuse
JSON: https://digestai.news/story/researchers-compose-frozen-rwkv-and-pythia-models-via-shared-latent-ad.json