In both Transformers and SSMs, in-context retrieval is performed by a small set of Gather and Aggregate heads, and the Transformer-SSM gap can be reduced to how well these few heads implement this mechanism.
Birdie: Advancing State Space Models with Reward-Driven Objectives and Curricula
1 Pith paper cite this work. Polarity classification is still indexing.
abstract
Efficient state space models (SSMs), such as linear recurrent neural networks and linear attention variants, offer computational advantages over Transformers but struggle with tasks requiring long-range in-context retrieval-like text copying, associative recall, and question answering over long contexts. Previous efforts to address these challenges have focused on architectural modifications, often reintroducing computational inefficiencies. In this paper, we propose a novel training procedure, Birdie, that significantly enhances the in-context retrieval capabilities of SSMs without altering their architecture. Our approach combines bidirectional input processing with dynamic mixtures of specialized pre-training objectives, optimized via reinforcement learning. We introduce a new bidirectional SSM architecture that seamlessly transitions from bidirectional context processing to causal generation. Experimental evaluations demonstrate that Birdie markedly improves performance on retrieval-intensive tasks such as multi-number phone book lookup, long paragraph question-answering, and infilling. This narrows the performance gap with Transformers, while retaining computational efficiency. Our findings highlight the importance of training procedures in leveraging the fixed-state capacity of SSMs, offering a new direction to advance their capabilities. All code and pre-trained models are available at https://www.github.com/samblouir/birdie, with support for JAX and PyTorch.
fields
cs.LG 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Understanding the Skill Gap in Recurrent Language Models: The Role of the Gather-and-Aggregate Mechanism
In both Transformers and SSMs, in-context retrieval is performed by a small set of Gather and Aggregate heads, and the Transformer-SSM gap can be reduced to how well these few heads implement this mechanism.