Code reasoning is essential for building AI systems that can understand, generate, and verify complex software. Yet most open-source code corpora focus on isolated snippets or single functions, failing to capture the long-range dependencies and cross-file interactions that define real-world repositories. We introduce MegaCode, a 300B repository-level corpus designed to support structured code reasoning and long-context modeling. MegaCode preserves full project structure—including build scripts, configuration files, and inter-file links—and is constructed through a multi-stage curation pipeline combining deduplication, rule-based filtering, and model-based quality assessment. Models pretrained on MegaCode demonstrate consistent gains on code reasoning and long-context benchmarks compared with those trained on flattened code datasets. Also, we validate the effectiveness of adding repo relevent infomation such as LSP, repo structure and repo summary. These results highlight the importance of maintaining repository structure for next-generation code intelligence. We release the dataset, curation pipeline, and evaluation suite to facilitate research on large-scale code pretraining and repository-level understanding.

Combinatorial optimization has long been dominated by manually engineered heuristics, a paradigm requiring substantial expert intuition and implementation overhead. The advent of Large Language Models has disrupted this landscape, enabling the autonomous synthesis and optimization of algorithms. Recent approaches typically iterate on heuristic populations using LLMs as mutators; however, these strategies often suffer from limited exploration, leading to stagnation in local optima. To overcome this, we present the Experience-Driven Reflective Co-Evolution of Prompt and Heuristics (EvoPH) for autonomous algorithm design, a novel framework that couples an island migration model with elite selection to maintain population diversity. Uniquely, EvoPH co-evolves both the guiding prompts and the heuristics themselves, using a feedback loop driven by past experience to refine the search process. We demonstrate EvoPH’s efficacy on the Traveling Salesman and Bin Packing Problems. Our results show that EvoPH achieves superior accuracy compared to baselines, marking a significant step forward in LLM-aided algorithm design.

Recent advances in large language models (LLMs) have enabled web agents to perform interactive tasks on real-world websites. However, existing agents still suffer from limited robustness, efficiency, and task success, largely due to their lack of structural understanding of websites and the absence of browsing priors in pre-trained models. To address these challenges, this paper proposes the Web Agent Sitemap Protocol (WASP), an agent-oriented sitemap that integrate structured website knowledge into web agents. WASP adopts a dual-granularity design, providing global site-level structure and local page-level semantic and interaction guidance. We also introduce a framework LightASM for constructing such sitemaps by identifying core pages and generating concise semantic summaries and block-level descriptions. Experiments on real-world browsing benchmarks demonstrate that WASP substantially improves the robustness, efficiency, and effectiveness of LLM-based web agents without extra training.

As access to high-quality, domain-specific data grows increasingly scarce, multi-epoch training has become a practical strategy for adapting large language models (LLMs). However, autoregressive models often suffer from performance degradation under repeated data exposure, where overfitting leads to a marked decline in model capability. Through empirical analysis, we trace this degradation to an imbalance in learning dynamics: predictable, low-entropy tokens are learned quickly and come to dominate optimization, while the model’s ability to generalize on high-entropy tokens deteriorates with continued training. To address this, we introduce EntroDrop, an entropy-guided token dropout method that functions as structured data regularization. EntroDrop selectively masks low-entropy tokens during training and employs a curriculum schedule to adjust regularization strength in alignment with training progress. Experiments across model scales from 0.6B to 8B parameters show that EntroDrop consistently outperforms standard regularization baselines and maintains robust performance throughout extended multi-epoch training. These findings underscore the importance of aligning regularization with token-level learning dynamics when training on limited data. Our approach offers a promising pathway toward more effective adaptation of LLMs in data-constrained domains.

Diffusion large language models (dLLMs) offer an efficient alternative to autoregressive models through parallel decoding, yet existing post-training methods largely rely on random masking strategies that overlook intrinsic token dependencies. In this work, we present an empirical analysis of attention in dLLMs and show that tokens attending more strongly to revealed context exhibit greater generation stability and play a critical role in reasoning. Motivated by these findings, we propose AGDO, an attention-guided denoising and optimization framework that aligns both training and optimization with attention-derived dependencies. AGDO determines the denoising order based on attention structure and emphasizes attention-critical tokens during supervised fine-tuning and reinforcement learning. Experiments on mathematical and coding benchmarks demonstrate that AGDO consistently improves reasoning performance, outperforming state-of-the-art post-training methods for dLLMs.