Search papers, labs, and topics across Lattice.
The paper introduces Self-Reflective Language Models (SRLM), a framework that enhances Recursive Language Models (RLM) by incorporating uncertainty-aware self-reflection to improve context-interaction program selection. SRLM uses self-consistency, reasoning length, and verbalized confidence as signals of internal uncertainty to evaluate candidate programs. Experiments across diverse benchmarks demonstrate that SRLM outperforms state-of-the-art baselines, achieving up to a 22% improvement over RLM, and showing that recursion is not the primary driver of RLM performance.
Forget fancy recursion: uncertainty-aware self-reflection alone can boost long-context language model performance by up to 22%, even surpassing Recursive Language Models (RLM).
Long-context handling remains a core challenge for language models: even with extended context windows, models often fail to reliably extract, reason over, and use the information across long contexts. Recent works like Recursive Language Models (RLM) have approached this challenge by agentic way of decomposing long contexts into recursive sub-calls through programmatic interaction at inference. While promising, the success of RLM critically depends on how these context-interaction programs are selected, which has remained largely unexplored. In this paper, we study this problem and introduce SRLM, a framework that augments programmatic context interaction with uncertainty-aware Self-Reflection. SRLM leverages three intrinsic signals: self consistency, reasoning length, and verbalized confidence. These serve as complementary indicators of a model's internal uncertainty, and the model uses them to evaluate and compare candidate context-interaction programs. Extensive experiments across diverse benchmark datasets, context lengths, and backbone models, show that SRLM consistently outperforms state-of-the-art baselines, yielding up to 22% improvement over RLM under the same time budget. Our findings show that recursion itself is not the primary driver of performance in RLM, and a simple self-reflective program search can match or surpass RLM without requiring self-query or explicit recursion mechanisms. We find that for context lengths within the model's window, RLMs with recursion often degrade performance relative to the base model, whereas SRLM yields consistent gains across both short and long contexts. We also find that RLM is less effective in tasks with semantically intensive nature, where heuristic program search is insufficient and broader contextual understanding is required, while self-reflection in SRLM provides a semantic signal that better steers reasoning in these scenarios.
