Giovanni Monea

⚠️ Stop pretraining your looped Transformer with a separate KV cache per recursion! In our recent preprint, we show that *memory sharing* not only saves memory but is also a net-positive inductive bias! Less memory, same flops, higher quality. 🔗 arxiv.org/abs/2610.02383 🧵

So, what does memory sharing bring to the table to *improve* quality? Prediction recursions attend heavily to the shared memory. Each read sends a gradient signal back to the state recursion. We find that this "gradient highway" makes the state recursion learn much more.

⚠️ Stop pretraining your looped Transformer with a separate KV cache per recursion! In our recent preprint, we show that *memory sharing* not only saves memory but is also a net-positive inductive bias! Less memory, same flops, higher quality. 🔗 arxiv.org/abs/2610.02383 🧵

Prior work laid the foundation for looped models and reduced their extra memory at inference. The result? Surprisingly strong despite the train-test mismatch. That made us ask: could this be a good inductive bias? One recursion memorizes, the others reuse it to predict.

⚠️ Stop pretraining your looped Transformer with a separate KV cache per recursion! In our recent preprint, we show that *memory sharing* not only saves memory but is also a net-positive inductive bias! Less memory, same flops, higher quality. 🔗 arxiv.org/abs/2610.02383 🧵

Sharing memory makes looping cheaper and better. At 1B with 5 recursions, H-LPT beats a same-size Transformer by 1.1 perplexity with 76% less memory. Locally, it's easier to fit on your GPU. For serving, it fits 3.6x bigger batches and keeps 81% of the Transformer's throughput.