Brief

Several recent research papers explore methods to enhance the reasoning capabilities of large language models (LLMs). One study suggests that increasing a model's long-context capacity improves reasoning performance across various tasks. Another paper introduces OckBench, a benchmark focused on measuring the token efficiency of LLM reasoning, highlighting significant room for optimization. Additional research proposes frameworks for evaluating inductive reasoning, improving robustness through invariant gradient alignment, and enabling belief-aware reasoning in multimodal models. AI

Multiple research papers released on arXiv explore advancements in AI agents, focusing on improving their reasoning, memory, and training efficiency. Qwen3.6-35B-A3B, an open-source sparse MoE model, demonstrates strong agentic coding capabilities. Other studies introduce methods for better skill presentation, long-context reasoning through RL, skill reuse as compression, and adaptive context management for agents tackling complex, long-horizon tasks. Additionally, research presents AutoSci, a system for automating the scientific research lifecycle, and PithTrain, a compact training framework for MoE models designed for agent-native development. AI

Researchers have developed several new methods to accelerate large language model (LLM) inference through speculative decoding. AdaPLD improves retrieval and draft construction by using semantic similarity and branched hypotheses, achieving up to 3.10x speedup. SSSD combines n-gram matching with hardware-aware speculation for up to 2.9x latency reduction without training. D^2SD uses a dual diffusion model and confidence-guided prefix trees to enhance acceptance rates, while TAPS optimizes prefix tree selection for diffusion-drafted decoding, yielding up to 7.9x speedup. KnapSpec treats draft model selection as a knapsack problem to maximize throughput, achieving up to 1.47x speedup, and Vegas uses verification-guided sparse attention for improved decoding throughput. Additionally, LK Losses directly optimize the acceptance rate during training, leading to gains of 8-10% in average acceptance length. AI