Pruning is a critical technique for compressing Large Language Models (LLMs), yet its effectiveness is highly sensitive to the small calibration set used for parameter importance estimation. Current selection relies on simple heuristics like random sampling or entropy, causing large variation in post-pruning perplexity. In this work, we reveal the source of this instability at two levels: calibration samples are not equally important, and a distinct quality hierarchy, comprising Golden, Mediocre, and Detrimental samples, exists within any data pool. More importantly, we demonstrate that this hierarchy is not fixed: the utility of a sample is highly context-dependent, varying with the specific model and pruning algorithm, which renders static filtering approaches fundamentally insufficient. To address these, we introduce OASIS, the first end-to-end framework that directly optimizes calibration data selection with respect to the pruned model’s downstream performance. OASIS leverages a differentiable soft-mask proxy to propagate task-level gradients back to data importance weights. Comprehensive experiments across the Llama and Qwen families show that OASIS improves calibration robustness across the evaluated pruning settings, establishing a reliable standard for data-aware LLM compression.

Retrieval-Augmented Generation (RAG) enhances Large Language Models (LLMs) by integrating them with an external knowledge base to improve the answer relevance and accuracy. In real-world scenarios, beyond pure text, a substantial amount of knowledge is stored in tables, and user questions often require retrieving answers that are distributed across multiple tables. Retrieving knowledge from a table corpora (i.e., various individual tables) for a question remains nascent, for (i) how to understand intra- and inter-table knowledge effectively, (ii) how to filter unnecessary tables and retrieve the most relevant tables efficiently, (iii) how to organize complex retrieved contexts for LLMs’ reasoning, and (iv) how to evaluate the corresponding performance in a realistic setting. Facing the above challenges, in this paper, we first propose a table-corpora-aware RAG framework, named T-RAG, which consists of the hierarchical memory index, multi-stage retrieval, and graph-aware context organization for effective and efficient table knowledge retrieval and inference. Then, we develop a multi-table question answering benchmark named MultiTableQA, which spans 3 different task types, 57,193 tables, and 23,758 questions in total, and the sources are all from real-world scenarios. Based on MultiTableQA, we perform a comprehensive comparison of table retrieval methods, RAG-based approaches, and table-to-graph representation learning methods. T-RAG consistently achieves state-of-the-art accuracy, recall, and runtime performance, with improvements of up to 9.4%. Moreover, T-RAG yields an average inference gain of 11.8% across different downstream backbone LLMs. Our code and data are available at https://github.com/jiaruzouu/T-RAG.

Diffusion language models (DLMs) have emerged as a promising alternative to the auto-regressive paradigm. With bidirectional attention and any-order generation, DLMs naturally fit infilling tasks, which require generating a middle span conditioned on both the prefix and the suffix. However, infilling is sensitive to the length of the span, while DLMs require the length to be fixed before generation. Although prior studies extend DLMs to dynamic lengths, they still suffer from two limitations. (i) *Sensitivity to initial length.* These methods require a preset length to initialize the search and are highly sensitive to this initial length, often yielding suboptimal results. (ii) *Inference inefficiency.* They either insert length-changing operations during generation or repeatedly search for an appropriate length using multi-step denoising confidence, both of which introduce substantial additional forward passes and computational cost. Therefore, we propose PILL (**P**robing-based **I**nfi**L**ing with preset-**L**ength-free decoding), an efficient infilling method for DLMs that requires no preset initial length and drastically reduces the number of forward passes compared to baselines. Experiments show that, across five DLMs spanning different families, architectures, and training recipes on eight infilling benchmarks, PILL improves over the strongest baseline by +4.8 average pass rate on code and +6.0 BLEU-2 on text, while running 1.82× faster than that baseline.

Reinforcement Learning with Verifiable Rewards (RLVR) has significantly advanced the reasoning capabilities of Large Language Models (LLMs). However, methods such as GRPO and DAPO suffer from substantial computational cost, since they rely on sampling many rollouts for each prompt. Moreover, in RLVR the relative advantage is often sparse: many samples become nearly all-correct or all-incorrect, yielding low within-group reward variance and thus weak learning signals. In this paper, we introduce ARRoL (Accelerating RLVR via online RoLlout Pruning), an online rollout pruning method that prunes rollouts during generation while explicitly steering the surviving ones more correctness-balanced to enhance learning signals. Specifically, ARRoL trains a lightweight quality head on-the-fly to predict the success probability of partial rollouts and uses it to make early pruning decisions. The learned quality head can further weigh candidates to improve inference accuracy during test-time voting. To improve efficiency, we present a system design that prunes rollouts inside the inference engine and re-batches the remaining ones for log-probability computation and policy updates. Across GRPO and DAPO on Qwen-3 and LLaMA-3.2 models (1B-8B), ARRoL improves average accuracy by +2.30 to +2.99 while achieving up to 1.7× training speedup, and yielding up to +8.33 additional gains in average accuracy in test-time voting.