AI Insight
LongZero is a new method that enables large language models to improve their long-context reasoning abilities without requiring any external training data. The approach works by having the model generate its own multi-hop reasoning questions and answers in a compact form, then expanding them into long multi-document contexts, which are then used for reinforcement learning training. Testing on two LLMs across five benchmarks showed that LongZero outperformed both baseline methods and training with externally-sourced data, demonstrating effective generalization to real-world tasks.
Why it matters
This research addresses a critical bottleneck in AI development: the need for expensive, high-quality training data to improve advanced reasoning capabilities. By enabling models to self-generate training materials, LongZero could accelerate progress in long-context understanding without dependence on human annotation or external corpora, particularly valuable as models become more capable and suitable training data becomes scarcer.
Understand the Science
⚠️ Preprint – Noch nicht peer-reviewed
Dieser Artikel wurde noch nicht von unabhängigen Experten begutachtet. Die Ergebnisse sind vorläufig und sollten mit Vorsicht interpretiert werden.
Abstract: Reinforcement learning (RL) is an effective way to strengthen the long-context reasoning of large language models (LLMs), but it needs training data that are long, diverse, challenging, and verifiable. Human annotation of such data is expensive, naturally occurring data are too scarce to sustain RL, and existing synthesis methods depend on external corpora or stronger teacher models, which become harder to obtain as the trained models get stronger. Recently, RL from zero data, in which a model learns from tasks it proposes itself, has achieved excellent performance, but existing methods target short-context tasks. Therefore, we propose $textbf{LongZero}$, a method that lets an LLM improve its own long-context reasoning without any external data. Instead of writing questions and answers over existing documents, LongZero first has the model synthesize the question and answer with its supporting evidence in a compact structured form, and only then separately expands the evidence into a long multi-document context. LongZero combines these self-written multi-hop reasoning tasks with self-written multi-round retrieval tasks and trains the model with GRPO. We evaluate LongZero on two LLMs across five long-context reasoning benchmarks and compare it with RL on data synthesized from external corpora. Experimental results show that LongZero can generalize to real tasks, and outperforms both baselines, suggesting a viable and promising path to improving long-context reasoning of LLMs.
Source: LongZero: Enhancing Long-Context Reasoning in LLMs from Zero Data