Rishabh

Xingyue Huang, Rishabh, Gregor Franke et al.

Recent advances in Large Language Models (LLMs) have shown that their reasoning capabilities can be significantly improved through Reinforcement Learning with Verifiable Reward (RLVR), particularly in domains like mathematics and programming, where ground-truth correctness can be automatically evaluated. However, extending this success to other reasoning-intensive domains remains challenging due to the scarcity of high-quality, verifiable datasets and the high cost of human supervision. In this work, we introduce the Loong Project: an open-source framework for scalable synthetic data generation and verification across a diverse range of reasoning-intensive domains. The framework consists of two key components: (1) LoongBench, a curated seed dataset containing 8,729 human-vetted examples across 12 domains (e.g., Advanced Mathematics, Chemistry, Logic), each paired with executable code and rich metadata; and (2) LoongEnv, a modular synthetic data generation environment that supports multiple prompting strategies to produce new question-answer-code triples. Together, these components form an agent-environment loop that enables reinforcement learning, where an LLM-based agent is rewarded for generating Chain-of-Thought (CoT) solutions that align with code-executed answers. Empirically, we benchmark LoongBench on a broad suite of both open-source and proprietary LLMs to evaluate domain coverage and reveal performance bottlenecks. In addition, we conduct a comprehensive analysis of synthetic data generated by LoongEnv, examining correctness, difficulty, and diversity. Code and documentation are available at https://github.com/camel-ai/loong.

Xingyue Huang, Xianglong Hu, Zifeng Ding et al.

Large language models (LLMs) often struggle with mathematical problems that require exact computation or multi-step algebraic reasoning. Tool-integrated reasoning (TIR) offers a promising solution by leveraging external tools such as code interpreters to ensure correctness, but it introduces inference-time dependencies that hinder scalability and deployment. In this work, we propose a new paradigm for distilling tool knowledge into LLMs purely through natural language. We first construct a Solver Agent that solves math problems by interleaving planning, symbolic tool calls, and reflective reasoning. Then, using a back-translation pipeline powered by multiple LLM-based agents, we convert interleaved TIR traces into natural language reasoning traces. A Translator Agent generates explanations for individual tool calls, while a Rephrase Agent merges them into a fluent and globally coherent narrative. Empirically, we show that fine-tuning a small open-source model on these synthesized traces enables it to internalize both tool knowledge and structured reasoning patterns, yielding gains on competition-level math benchmarks without requiring tool access at inference.

Piyus Kedia, Rahul Purandare, Udit Kumar Agarwal et al.

Spatial safety violations are the root cause of many security attacks and unexpected behavior of applications. Existing techniques to enforce spatial safety work broadly at either object or pointer granularity. Object-based approaches tend to incur high CPU overheads, whereas pointer-based approaches incur both high CPU and memory overheads. SGXBounds, an object-based approach, is so far the most efficient technique that provides complete out-of-bounds protection for objects. However, a major drawback of this approach is that it can't support address space larger than 32-bit. In this paper, we present CGuard, a tool that provides object-bounds protection for C applications with comparable overheads to SGXBounds without restricting the application address space. CGuard stores the bounds information just before the base address of an object and encodes the relative offset of the base address in the spare bits of the virtual address available in x86_64 architecture. For an object that can't fit in the spare bits, CGuard uses a custom memory layout that enables it to find the base address of the object in just one memory access. Our study revealed spatial safety violations in the gcc and x264 benchmarks from the SPEC CPU2017 benchmark suite and the string_match benchmark from the Phoenix benchmark suite. The execution time overheads for the SPEC CPU2017 and Phoenix benchmark suites were 42% and 26% respectively, whereas the reduction in the throughput for the Apache webserver when the CPUs were fully saturated was 30%. These results indicate that CGuard can be highly effective while maintaining a reasonable degree of efficiency.