Long-form narratives are challenging for long-context models because their structure is implicit: events, characters, and plotlines interact across hundreds of pages without the explicit cues that guide navigation in structured documents. We address this by constructing storyline trees, hierarchical representations that organize narratives from global themes and major plotlines to fine-grained events. We first segment chapters into contiguous narrative segments, or scenes, and use them as the basic units for tree construction. We then infer storyline trees through complementary top-down and bottom-up procedures that derive, refine, cluster, and summarize storylines at multiple levels of abstraction. We showcase the utility of this representation for question answering: storyline trees enable adaptive retrieval, allowing models to iteratively inspect high-level narrative structure and retrieve scene-level evidence on demand. Experiments on three long-context narrative QA benchmarks show that adaptive retrieval outperforms strong baselines, including post-trained long-context models and agentic chunk-based methods. Ablations confirm that scenes are more effective basic units than chapters or generic segmentation, and that gains persist under matched retrieval budgets.

Confidence in LLMs is a useful indicator of model uncertainty and answer reliability. Existing work mainly focused on single-turn scenarios, while research on confidence in complex multi-turn interactions is limited. In this paper, we investigate whether LLM-based search agents have the ability to communicate their own confidence through verbalized confidence scores after long sequences of actions, a significantly more challenging task compared to outputting confidence in a single interaction. Experimenting on open-source agentic models, we first find that models exhibit much higher task accuracy at high confidence while having near-zero accuracy when confidence is low. Based on this observation, we propose Test-Time Scaling (TTS) methods that use confidence scores to determine answer quality, encourage the model to try again until reaching a satisfactory confidence level. Results show that our proposed methods significantly reduce token consumption while demonstrating competitive performance compared to baseline fixed budget TTS methods.

Modern instruction-tuning mixtures are increasingly large, redundant, and heterogeneous, making full-data supervised fine-tuning costly and often unnecessary. We propose GRADFILTERING, an objective-agnostic, validation-free data selection method that scores examples using uncertainty-aware proxy training dynamics. GRADFILTERING trains a lightweight proxy model with a LoRA ensemble and ranks examples by a Gradient Signal-to-Noise Ratio (G-SNR), which combines early-to-late gradient drop with late-stage ensemble disagreement. Because selection requires neither target validation examples nor target-model gradients, the selected subsets can be reused across downstream targets. Across Alpaca and Tulu-3-52K, dense LLaMA-2-7B, LLaMA-3-8B, and Qwen3-8B targets, and a Qwen3-30B-A3B MoE target, GRADFILTERING-selected 5–15% subsets consistently outperform random selection and are competitive with the target-aware LESS-T reference while remaining validation-free. We further evaluate broad model capabilities on objective benchmarks and analyze proxy choice, gradient windows, ensemble size, human preferences, and selection cost, demonstrating the practical utility of early proxy dynamics for reusable, compact instruction-tuning data selection. Code is available at this GitHub repository.

Hierarchical Merging is a technique commonly used to summarize very long texts (>100K tokens) by breaking down the input into smaller sections, summarizing those sections individually, and then merging or combining those summaries into a final coherent summary. Although it helps address the limitations of large language models (LLMs) with fixed input length constraints, the recursive merging process can amplify LLM hallucinations, increasing the risk of factual inaccuracies. In this paper, we seek to mitigate hallucinations by enriching hierarchical merging with context from the source document. Specifically, we propose different approaches to contextual augmentation ranging from replacing intermediate summaries with relevant input context, to refining them while using the context as supporting evidence, and aligning them implicitly (via citations) to the input. Experimental results on datasets representing legal and narrative domains show that contextual augmentation consistently outperforms zero-shot and hierarchical merging baselines for the Llama 3.1 model family. Our analysis further reveals that refinement methods tend to perform best when paired with extractive summarization for identifying relevant input.