This paper investigates how different types of compressed reasoning data (Explicit, Composed, Implicit CoT) affect LLM performance during post-training, finding that the choice of compression and subs…

The paper introduces Hierarchical Adaptive Budgeter (HAB), a framework that improves LLM reasoning efficiency by adaptively allocating computational resources to match the intrinsic complexity of both…

This paper proposes CoLT, a framework that enables multi-modal models to reason through a chain of latent thought representations instead of text tokens, improving performance and reducing inference t…

The paper introduces CosmicFish-HRM, a compact language model that achieves adaptive reasoning by dynamically allocating computational effort through a Hierarchical Reasoning Module (HRM), showing tha…

DenseSteer is a training-free inference-time framework that improves the math reasoning capabilities of small language models by steering their internal representations toward a 'Dense Reasoning' patt…

The paper introduces Thinking as Compression (TaC), a novel paradigm showing that the inherent reasoning process of a large language model can naturally compress long context inputs, outperforming ded…

The paper introduces a multilingual benchmark (MentalMap) to test if LLMs build internal spatial world models from text, finding a universal 'L3 reasoning cliff' suggesting that text-only working memo…

This paper proposes X$^3$-OPD, a framework for transferring reasoning capabilities from text-based models to audio-language models using on-policy distillation.

This paper localizes the attention heads within LLMs responsible for specific reasoning steps, finding that specialized heads handle factual retrieval while higher layers manage global information int…

The paper evaluates LLM reasoning on Boolean satisfiability (SAT) problems, concluding that conventional metrics are misleading and proposing a paired-formula protocol with Accurate Differentiation Ra…

The paper proposes CSMR, a cognitive scheduling framework that allows a language model to dynamically decide when to acquire task-relevant visual evidence, significantly improving multimodal reasoning…

This paper identifies the root cause of performance degradation in full-duplex Spoken Language Models (SLMs) due to modality interference and proposes Lychee-FD, a framework that decouples conflicting…

The paper introduces Reasoning in Memory (RiM), a latent reasoning method that replaces autoregressive token generation with fixed memory blocks to enable compute-efficient internal working memory for…

The paper introduces OCC-RAG, a family of compact, task-specialized Small Language Models (SLMs) designed to achieve highly faithful, multi-hop question answering grounded strictly in provided context…

The paper proposes SLAT, a segment-level adaptive trimming framework, which efficiently reduces redundant reasoning in large language model CoT outputs by selectively suppressing segments with low mar…

This paper proposes a training-free framework called ReasonAlloc to mitigate inference bottlenecks in large language models by recasting decoding-time key-value compression as a hierarchical budget al…

This paper proposes a training-free framework called ReasonAlloc to mitigate inference bottlenecks in large language models by recasting decoding-time key-value compression as a hierarchical budget al…

This paper proposes an efficient human intervention mechanism, Deep Interaction, for correcting reasoning errors in large language models, achieving over 25% improvement in correction success rate and…