PULK!T

Transformers are data‑hungry in sequential tasks because they lack the right inductive bias. It’s well known that for many sequential problems (from adding numbers to step‑by‑step agentic execution and multi‑hop reasoning), transformers fail to generalize to longer sequences than they were trained on. “Train short, test long” often fails. The usual workaround is to "just train on whatever length you’ll need at test time". --------- 📉 But we show the consequence of this is data inefficiency: • Transformers can learn tasks for a single fixed sequence length fairly efficiently, but learning across multiple lengths requires much more data. • More importantly, transformers tend not to share mechanisms across tasks of different lengths; instead, they often learn isolated, length‑specific solutions. --------- 🧪 A simple way to test this: Consider modular addition (with and without CoT). Train a model to add 2, 3, …, L numbers at once and measure the data needed. Then train separate models for each length (2, 3, …, L) and sum their data requirements. 💡The intuition: If a model truly shares mechanisms across lengths, learning a distribution of lengths should require far fewer samples than learning each length separately. This comes from amortizing the learning cost: data for length n also helps the model learn length n+k. --------- 📊 Results: Sharing Factor κ = (sum of samples to learn each length separately) ÷ (samples to learn all lengths jointly) - κ > 1: mechanism sharing and amortized learning. - κ ≈ 1: learning length-specific solutions in isolation. - κ < 1: destructive interference; length-specific solutions compete for model capacity. Transformers showed low sharing factors, and even destructive interference with CoT. --------- ✨ Implications: This suggests that end-to-end learning in applied agentic settings, like robotics or GUI control, could be even more challenging. If data requirements grow unfavorably with sequence length, that might also help explain the persistent issues we see at large context lengths (e.g., context rot). Standard attention mechanism appears inefficient for step-by-step tasks, and we may ultimately be better off with recurrent agents.

1/N. Modern linear RNNs miss one crucial capability from classic RNNs: state-tracking. State-tracking is the ability to dynamically maintain an internal state of a system. It's essential for complex reasoning tasks, such as: - Updating a chessboard state after a sequence of moves. - Tracking variable values while executing code line-by-line. - Following an algorithm sequentially by keeping track of its intermediate state. These tasks demand more than pattern recognition; they require a persistent, updatable internal state. Our new paper investigates the architectural properties that enable this in linear RNNs. We introduce a taxonomy of these models, creating a hierarchy based on their fundamental state-tracking capabilities: