Recent Posts

Agentic Societies Need a Social Harness

智能体社会需要一个社会性缰绳

9/17/2026

arxivaics.ma

Coupled Calibration and Learning: Mitigating Teacher Bias in LLM Distillation without Target-Domain Reward Feedback

耦合的校准与学习：无目标域奖励反馈时缓解蒸馏中的教师偏差

9/17/2026

arxivaics.lg

Decomposition Buys Integrity, Not Yield

分解买的是完整性，不是产出

9/17/2026

arxivaics.ma

JustFit: 200K-Token LLM Serving on a 24 GiB Laptop with Just-in-Time State Management

JustFit：用即时状态管理在 24 GiB 笔记本上跑 20 万 token

9/17/2026

arxivaics.ai

When Should LLMs Abstain? Chain-of-Self-Questioning for Selective Risk Control

大模型何时该弃答？用自我追问链做选择性风险控制

9/17/2026

arxivaics.cl

Attention Mean Fields Predict Average Representation Dynamics

注意力均值场能预测表征的平均演化

9/16/2026

arxivaics.lg

Implementing a White-Box Undetectable Backdoor for Random Fourier Features

给随机傅里叶特征实现白盒不可检测后门

9/16/2026

arxivaics.cr

Reasoning with Image Generation

用图像生成来推理

9/16/2026

arxivaics.cv

Register Tokens for Bounded-State Reasoning in Diffusion Language Models

用寄存器 token 给扩散语言模型做定长状态推理

9/16/2026

arxivaics.cl

Where Post-Training Quantization Breaks Text Embedders

训练后量化在文本嵌入模型上会从哪里断

9/16/2026

arxivaics.ir

AGENTQ: A Checkpoint Can Pass Every Audit and Misbehave Once Quantized

AGENTQ：一个检查点可以通过所有审计，却在量化后行为不端

9/15/2026

arxivaics.cr

Measuring the Creativity of Frontier LLMs in Automated Research

度量前沿大模型在自动化科研中的「创造力」

9/15/2026

arxivaics.cl