Test-time scaling improves the inference performance of Large Language Models (LLMs) but also incurs substantial computational costs. Although recent studies have reduced token consumption through dynamic self-consistency, they remain constrained by the high latency of sequential requests. In this paper, we propose SeerSC, a dynamic self-consistency framework that simultaneously improves token efficiency and latency by integrating System 1 and System 2 reasoning. Specifically, we utilize the rapid System 1 to compute the answer entropy for given queries. This score is then used to evaluate the potential of samples for scaling, enabling dynamic self-consistency under System 2. Benefiting from the advance and accurate estimation provided by System 1, the proposed method can reduce token usage while simultaneously achieving a significant decrease in latency through parallel generation. It outperforms existing methods, achieving up to a 47% reduction in token consumption and a 43% reduction in inference latency without significant performance loss.

Large language models (LLMs) rely on key-value cache (KV cache) to accelerate decoding by reducing redundant computations. However, the KV cache memory usage grows substantially with longer text sequences, posing challenges for efficient deployment. Existing KV cache eviction methods prune tokens using prefilling-stage attention scores, causing inconsistency with actual inference queries, especially under tight memory budgets. In this paper, we propose Lookahead Q-Cache (LAQ), a novel eviction framework that generates low-cost pseudo lookahead queries to better approximate the true decoding-stage queries. By using these lookahead queries as the observation window for importance estimation, LAQ achieves more consistent and accurate KV cache eviction aligned with real inference scenarios. Experimental results on LongBench and Needle-in-a-Haystack benchmarks show that LAQ outperforms existing methods across various budget levels, achieving a 1 4 point improvement on LongBench under limited cache budget. Moreover, LAQ is complementary to existing approaches and can be flexibly combined to yield further improvements.

Nowadays, data augmentation through synthetic data has been widely used in the field of Grammatical Error Correction (GEC) to alleviate the problem of data scarcity. However, these synthetic data are mainly used in the pre-training phase rather than the data-limited fine tuning phase due to inconsistent error distribution and noisy labels. In this paper, we propose a synthetic data construction method based on contextual augmentation, which can ensure an efficient augmentation of the original data with a more consistent error distribution. Specifically, we combine rule-based substitution with model-based generation, using the generation model to generate a richer context for the extracted error patterns. Besides, we also propose a relabeling-based data cleaning method to mitigate the effects of noisy labels in synthetic data. Experiments on CoNLL14 and BEA19-Test show that our proposed augmentation method consistently and substantially outperforms strong baselines and achieves the state-of-the-art level with only a few synthetic data.

Existing speculative decoding methods typically require additional model structure and training processes to assist the model for draft token generation. This makes the migration of acceleration methods to the new model more costly and more demanding on device memory. To address this problem, we propose the Make Some Noise (MSN) training framework as a replacement for the supervised fine-tuning stage of the large language model. The training method simply introduces some noise at the input for the model to learn the denoising task. It significantly enhances the parallel decoding capability of the model without affecting the original task capability. In addition, we propose a tree-based retrieval-augmented Jacobi (TR-Jacobi) decoding strategy to further improve the inference speed of MSN models. Experiments in both the general and code domains have shown that MSN can improve inference speed by 2.3-2.7x times without compromising model performance. The MSN model also achieves comparable acceleration ratios to the SOTA model with additional model structure on Spec-Bench.