Search papers, labs, and topics across Lattice.
This paper introduces Self-Guided Test-Time Training (S-TTT) to enhance long-context utilization in large language models (LLMs) by enabling models to identify and focus on relevant evidence spans during test-time adaptation. The authors demonstrate that traditional test-time training methods suffer from performance degradation when applied to randomly sampled spans, while their proposed method significantly improves accuracy by selectively training on oracle spans. Experiments on LongBench-v2 and LongBench-Pro show that S-TTT achieves up to a 15% relative improvement in accuracy for models like Qwen3-4B-Thinking-2507 and Llama-3.1-8B-Instruct.
Selecting relevant evidence spans for test-time training can boost long-context LLM accuracy by up to 15%.
Long-context processing has become increasingly important for large language models (LLMs), but simply extending the context window does not guarantee effective utilization of long inputs. As input length grows, accuracy often degrades, indicating that models still struggle to identify and use the evidence most relevant to a question. A promising way to improve long-context utilization is test-time training (TTT), which treats the test context as a training example for instance-specific parameter adaptation. However, applying TTT to the entire long context is prohibitively expensive, while adapting on randomly sampled spans introduces severe noise. Because most spans in a long context are irrelevant to the specific question, training on them may even degrade the base model's performance. Our preliminary study shows that TTT is highly sensitive to training-span quality: on LongBench-v2, TTT on randomly sampled spans hurts performance, whereas TTT on oracle spans substantially improves it. Motivated by this, we propose a simple method, Self-Guided TTT (S-TTT): before adaptation, the model identifies the evidence spans it should learn from, and the standard language-modeling training objective is applied only to those selected spans. On two challenging long-context reasoning benchmarks, LongBench-v2 and LongBench-Pro, S-TTT improves accuracy for both Qwen3-4B-Thinking-2507 and Llama-3.1-8B-Instruct, achieving up to a 15% relative improvement.
