论文

GradCuit（梯度穿越电路）在所选 Transformer 层、提示隐藏表示与生成续写之间插入可优化的潜变量，开启了鲁棒且可解释的测试时缩放新维度，使 LLM 调整其推理方式，而不仅仅是重新生成、采样或重排输出。GradCuit (gradient through circuit), which inserts optimizable latent states at a selected Transformer layer between the hidden representations of the prompt and the generated continuation, opens a new axis of robust and interpretable test-time scaling, where LLMs adapt how they reason rather than merely regenerate, sample, or rerank outputs.

本文提出 GPTQ-2D，以三次时间复杂度生成相同的取整矩阵，并研究该任务的双侧版本——固定非奇异基矩阵同时作用于残差的左右两侧This work presents GPTQ-2D, which produces the identical rounded matrix in cubic time, and studies the two-sided version of this task, in which fixed nonsingular basis matrices act on both the left and the right of the residual.

简化了 MoE 路由算法，设计出通信与计算成本更低的直观改进模型，并首次证明大型稀疏模型可以使用更低精度格式进行训练This work simplifies the MoE routing algorithm and design intuitive improved models with reduced communication and computational costs and shows large sparse models may be trained, for the first time, with lower precision formats.

Pointer sentinel-LSTM 模型在 Penn Treebank 上以远少于标准 softmax LSTM 的参数量达到 SOTA 语言建模性能，并开源了 WikiText 语料库The pointer sentinel-LSTM model achieves state of the art language modeling performance on the Penn Treebank while using far fewer parameters than a standard softmax LSTM and the freely available WikiText corpus is introduced.

DASH (Drift Aware advantage SHaping) 根据每个推理片段是朝向还是偏离正确性来分配片段级 credit，在 overthinking 普遍的场景下取得最高准确率，同时减少 overthinking 行为，比基线实现更高效的自纠错。DASH (Drift Aware advantage SHaping), which assigns segment-level credit based on whether each reasoning segment leads toward or away from correctness, achieves the highest accuracy where overthinking is prevalent while reducing overthinking behaviors and achieving more productive self-correction than baselines.

提出 ARCHead，一种紧凑型 LM-head 压缩器，结合量化低秩核心、组级 INT4 残差以及基于激活导出度量拟合的低秩校正，在 block 量化器无法触及的大输出投影上实现互补压缩。ARCHead is presented, a packed LM-head compressor that combines a quantized low-rank core, group-wise INT4 residuals, and a low-rank correction fitted in an activation-derived metric to complements block quantizers by compressing the large output projection they can leave untouched.

提出 RestoreKV，在相同 KV 总预算下以可学习的恢复机制补充基于选择的 KV 压缩方案，显著降低压缩带来的性能损失。RestoreKV is introduced, which complements this selection-based formulation with learned restoration with learned restoration under the same total KV budget, and substantially reduces compression-induced degradation.

在三种大语言模型上的实验表明，思维链提示能够在一系列算术、常识和符号推理任务上提升性能。Experiments on three large language models show that chain of thought prompting improves performance on a range of arithmetic, commonsense, and symbolic reasoning tasks.

本工作提出两种参数削减技术以降低 BERT 的内存占用并提升训练速度，并采用一种聚焦于建模句子间连贯性的自监督损失。This work presents two parameter-reduction techniques to lower memory consumption and increase the training speed of BERT, and uses a self-supervised loss that focuses on modeling inter-sentence coherence.

研究表明，利用变分方法最新进展的深度生成模型与近似贝叶斯推断能够带来显著提升，使生成式方法在半监督学习上极具竞争力。It is shown that deep generative models and approximate Bayesian inference exploiting recent advances in variational methods can be used to provide significant improvements, making generative approaches highly competitive for semi-supervised learning.

这项系统性研究在数十项语言理解任务上比较了预训练目标、架构、无标注数据集、迁移方法及其他因素，并在涵盖摘要、问答、文本分类等的许多基准上取得了 SOTA 结果。This systematic study compares pre-training objectives, architectures, unlabeled datasets, transfer approaches, and other factors on dozens of language understanding tasks and achieves state-of-the-art results on many benchmarks covering summarization, question answering, text classification, and more.

本文提出了 mixup，一种通过对样本对及其标签的凸组合来训练神经网络的简单学习原则，提升了 SOTA 神经网络架构的泛化能力。This work proposes mixup, a simple learning principle that trains a neural network on convex combinations of pairs of examples and their labels, which improves the generalization of state-of-the-art neural network architectures.

本文提出了一种双时间尺度更新规则（TTUR），用于在任意 GAN 损失函数下使用 SGD 训练 GAN，并引入了 Frechet Inception Distance（FID），相比 Inception Score 能更好地捕捉生成图像与真实图像之间的相似性。This work proposes a two time-scale update rule (TTUR) for training GANs with stochastic gradient descent on arbitrary GAN loss functions and introduces the "Frechet Inception Distance" (FID) which captures the similarity of generated images to real ones better than the Inception Score.

名为 PaLM 的 540 亿参数、密集激活的 Transformer 语言模型取得了突破性性能，在一系列多步推理任务上超越了微调后的 SOTA，并在最近发布的 BIG-bench 基准上超越了人类平均水平。A 540-billion parameter, densely activated, Transformer language model, which is called PaLM achieves breakthrough performance, outperforming the finetuned state-of-the-art on a suite of multi-step reasoning tasks, and outperforming average human performance on the recently released BIG-bench benchmark.

研究表明，APE 生成的提示词既可引导模型趋向真实性和/或信息量，也可通过将其前置拼接到标准上下文学习提示词之前来提升少样本学习性能。It is shown that APE-engineered prompts can be applied to steer models toward truthfulness and/or informativeness, as well as to improve few-shot learning performance by simply prepending them to standard in-context learning prompts.

HiSparse 已合入上游 SGLang，并在 H200、B200 与 GH200 平台上针对 DSA、NSA、Quest 三类稀疏注意力家族进行评估：在长上下文负载下峰值生成吞吐提升最高达 4.7×，同时保持可比的 per-token 延迟，并降低高负载下的 time-to-first-token。HiSparse is merged into upstream SGLang and evaluated across three sparse-attention families (DSA, NSA, and Quest) on H200, B200, and GH200 platforms: it improves peak generation throughput by up to 4.7x on long-context workloads while preserving comparable per-token latency and reducing time-to-first-token at high load.

论文证明，使用标注数据进行微调，并允许模型查询外部知识源，能够在安全性和事实性这两个关键挑战上带来显著提升。It is demonstrated that fine-tuning with annotated data and enabling the model to consult external knowledge sources can lead to significant improvements towards the two key challenges of safety and factual grounding.

研究表明 BigBird 是序列函数的通用逼近器，且具备图灵完备性，从而保留了二次全注意力模型的这些性质。It is shown that BigBird is a universal approximator of sequence functions and is Turing complete, thereby preserving these properties of the quadratic, full attention model.

本文提出 OasisKV，一种以显存为中心的 LLM 推理系统设计，通过在 LLM 解码期间将完整 KV-cache 存储与 HBM 解耦来缓解 HBM 容量压力，并观察到未来重要 token 可借助推测解码（SD）所起草的前瞻 token 被提前准确预测。OasisKV is presented, a memory-centric LLM inference system design that alleviates HBM capacity pressure by decoupling full KV-cache storage from HBM during LLM decoding and observes that future important tokens can be predicted accurately in advance using lookahead tokens drafted by speculative decoding (SD).

一个结构定理将无记忆等变规则刻画为恰好由 Gram 矩阵决定的左预处理子；一个迁移定理将梯度流的路径性质推广到 common-scalar 流。A structure theorem characterizes the memoryless equivariant rules as exactly the Gram-determined left preconditioners, and a transfer theorem carries gradient flow's pathwise properties to common-scalar flows.

本文讨论了一种被称为大语言模型涌现能力的不可预测现象——若某项能力在小模型中不存在而在大模型中存在，则称为涌现。This paper discusses an unpredictable phenomenon that is referred to as emergent abilities of large language models, an ability to be emergent if it is not present in smaller models but is present in larger models.

本工作研究图结构输入的特征学习技术，并在程序验证任务上取得 SOTA 性能，该任务需将子图与抽象数据结构进行匹配。This work studies feature learning techniques for graph-structured inputs and achieves state-of-the-art performance on a problem from program verification, in which subgraphs need to be matched to abstract data structures.

提出 MultiMedQA 基准，整合六个现有医学问答数据集（涵盖专业医学、研究与消费者查询）及一个全新的在线医学问题搜索数据集，并提出针对模型答案的人工评估框架，揭示了 LLM 在医学领域的潜在应用价值。MultiMedQA, a benchmark combining six existing medical question answering datasets spanning professional medicine, research and consumer queries and a new dataset of medical questions searched online, is presented and a human evaluation framework for model answers is proposed, suggesting the potential utility of LLMs in medicine.

本文给出 ICL 的形式化定义，厘清其与相关研究的联系，并梳理讨论训练策略、提示设计策略及相关分析等高级技术。This paper presents a formal definition of ICL and clarify its correlation to related studies, and organizes and discusses advanced techniques, including training strategies, prompt designing strategies, and related analysis.

iFAN 提出 Adjusted Probability-Mask Ranking (APMR)，将查询竞争与预测的掩码质量对齐，抑制高置信度但不准确的竞争者；同时采用 Cross-Layer Self-Distillation (CLSD) 将更强的中间预测传递至最终层。iFAN introduces Adjusted Probability-Mask Ranking (APMR), which aligns query competition with predicted mask quality and suppresses high-confidence but inaccurate competitors, and employs Cross-Layer Self-Distillation (CLSD) to transfer stronger intermediate predictions to the final layer.

一个仅依赖字符级输入的简单神经语言模型，仅从字符即可编码语义和正字法信息，表明在许多语言中，字符输入足以完成语言建模。A simple neural language model that relies only on character-level inputs that is able to encode, from characters only, both semantic and orthographic information and suggests that on many languages, character inputs are sufficient for language modeling.

描述了一种无监督学习通用分布式句子编码器的方法，利用书籍文本的连续性，训练编码器-解码器模型以重建编码段落的周围句子。The approach for unsupervised learning of a generic, distributed sentence encoder is described, using the continuity of text from books to train an encoder-decoder model that tries to reconstruct the surrounding sentences of an encoded passage.

本文提出 Reduced Matrix Multiplication，一种免训练的输入自适应推理方法，通过沿收缩维度选取信息性切片来减少 Transformer 矩阵乘积，且不修改模型权重，并表明同一原理可扩展到多模态视觉-语言推理。Reduced Matrix Multiplication is proposed, a training-free, input-adaptive inference method that reduces Transformer matrix products by selecting informative slices along their contraction dimensions, without modifying model weights, and it is shown that the same principle extends to multimodal vision-language inference.

论文表明，使用 Stanford Natural Language Inference 数据集有监督训练的通用句子表示，在广泛的迁移任务上能持续优于 SkipThought vectors 等无监督方法。It is shown how universal sentence representations trained using the supervised data of the Stanford Natural Language Inference datasets can consistently outperform unsupervised methods like SkipThought vectors on a wide range of transfer tasks.

提出 HPSE，构建一种混合 rollout，能够在 student 自身轨迹中 coverage 缺失的精确位置补齐缺失事实，同时在其他位置保持 on-policy。HPSE is proposed, which builds a hybrid rollout that steps in to place missing facts onto the student's own trajectory precisely where its coverage fails, while staying on-policy elsewhere.

Gambit 通过周期性剪除低潜力轨迹并立即从高质量前缀分支，利用轻量 scorer 探测 hidden state，将算力动态集中到最有潜力的推理路径上，同时保持持续的高硬件利用率。By periodically pruning unpromising trajectories and immediately branching from high-quality prefixes, Gambit dynamically concentrates compute onto the most promising reasoning traces via a light-weight scorer probing hidden states while maintaining continuous high hardware utilization.

通过跨 46 个任务（涵盖四个认知领域）的电路分析，发现 LLM 发展出与人类大脑相似的模块化架构：在人类大脑中依赖同一网络的任务会在 LLM 中招募重叠的神经元，而依赖不同网络的任务则招募不同的神经元。Using circuit analyses across N=46 tasks spanning four cognitive domains, it is found that Large Language Models develop a modular architecture that mirrors the human brain: tasks drawing on the same network in humans recruit overlapping neurons in LLMs, whereas tasks drawing on different networks recruit distinct neurons.

FreeToken 是一个 edge-native 的 MoE serving 系统，将个人机器视为一个统一、弹性的推理平台而非小型 GPU，把开放权重转化为可部署的本地软件，使用户已有的机器成为运行前沿规模智能的实用平台。FreeToken is an edge-native MoE serving system that treats a personal machine not as a small GPU, but as a unified, elastic inference platform, turning open weights into deployable local software, making the machines users already own a practical platform for frontier-scale intelligence.