WRITING / FIELD NOTES写作 / 研究手记

A place to think in public.把思考写出来,供人检验。

Short, concrete notes from generative modeling, multimodal learning, agents, and training systems. Each essay starts with a design question.来自生成模型、多模态学习、智能体与训练系统的短文,从一个具体的设计问题开始。

IDEASMETHODSENGINEERING
ALL NOTES全部文章08 / 2026
01 RESEARCH NOTE

When should a model stop thinking?模型什么时候该少想一步?

CATER routes each question by the student's current success: develop uncertain capability with RL, then compress reliable reasoning through concise self-distillation.CATER 依据学生当前解题情况分配训练信号:先用强化学习巩固不稳定的能力,再通过简洁自蒸馏压缩稳定成功的推理。

REASONING EFFICIENCYICLR 2027 · UNDER REVIEWSEP 2026
02 RESEARCH NOTE

When the teacher disagrees with a correct answer教师与正确解答意见相左时

SP-OPD projects the teacher target onto a constraint supplied by a verifier, then carries that target through both distillation loss terms.SP-OPD 将教师分布投影到验证器给出的约束内,并让两项蒸馏损失使用同一个目标分布。

ON-POLICY DISTILLATIONICLR 2027 · UNDER REVIEWSEP 2026
03 RESEARCH NOTE

The unit of video interaction is a trajectory流式视频评测,该看整段交互

SVI-Bench scores speaking and silence together, while InteractFlow preserves the video state and timing behind every response.SVI-Bench 联合评估开口与沉默;InteractFlow 保留每次回答对应的视频画面和时间。

STREAMING VIDEOICLR 2027 · UNDER REVIEWSEP 2026
04 RESEARCH NOTE

When does a color become the car's?颜色,究竟属于谁?

A diffusion model can draw every requested object and color, then assign the colors to the wrong objects. Counterfactual prompt switches offer a way to study when that assignment changes.扩散模型可能画对对象和颜色,却配错归属。反事实提示词切换提供了研究归属何时还能改变的方法。

GENERATIVE MODELSICLR 2027 · UNDER REVIEWSEP 2026
05 RESEARCH NOTE

A vector is not yet a token向量进了模型,不代表模型读懂了

Projecting fixed multimodal features to the right hidden size solves a wiring problem. Readability needs an interface—and controls that show what the backbone actually contributes.把固定多模态特征投影到正确维度,只解决了接线问题。可读性需要更好的接口,也需要辨明骨干贡献的对照实验。

MULTIMODAL LEARNINGSEP 2026
06 RESEARCH NOTE

A hard question is not necessarily a hard search难题,不一定需要难搜索

Multi-hop questions can be solved with one search. What matters for agent training is the path a model must take to find the evidence.多跳问题也可能一次检索就得到答案。训练搜索智能体时,更重要的是模型找到证据的路径。

SEARCH AGENTSSEP 2026
07 RESEARCH NOTE

A search agent needs a ledger, not a longer transcript搜索智能体需要一本账,而非更长的对话

A persistent record of evidence, candidates, and missing constraints helps the agent carry useful state through a long run.持久记录证据、候选与缺失约束,才能让智能体在长程任务里带着有效状态前进。

AGENT STATESEP 2026
08 ENGINEERING NOTE

The data contract is part of the model数据契约,也是模型的一部分

Supervision masks, packed sequences, tool turns, and checkpoint lineage decide which behavior the model can actually learn.监督掩码、序列打包、工具轮次和模型权重血缘,决定模型究竟学到了什么。

POST-TRAININGSEP 2026
Prefer a reader?习惯用阅读器?Subscribe by RSS订阅 RSS ↗