Tag
A laptop-scale replication of Balestriero & LeCun's "good views beat next-token prediction" claim using only human COCO captions and 11M-param text encoders confirms the ordering (good views 0.320 > bad views 0.192 > next-token 0.148 on mutual 10-NN with DINOv2), while noting smaller retrieval gaps and scale limitations.
The paper introduces a controlled evaluation framework separating unique class support from representation allocation to determine when encoder adaptation remains worthwhile after classifier correction in long-tailed recognition, showing benchmark construction alone can flip this training decision.
The paper frames continual reinforcement learning with a contextual MDP that splits context into behavioral and physical components, and studies zero-shot reuse of world models in FourRooms tasks with Dyna-Q and TD-MPC, finding that physical-context changes substantially degrade reuse depending on the observation representation. It identifies representation generalization as a key bottleneck for reliable world-model reuse in continual RL.
This paper evaluates frozen video representations and finds that latent prediction errors in video world models reflect appearance and motion rather than reliably measuring physical surprise, presenting a framework that separates representation content from error signals across IntPhys2, Physion++, and Kubric benchmarks.
Yann LeCun and Randall Balestriero mathematically prove the Platonic Representation Hypothesis and show that JEPA training with scene-related positive views achieves image-text alignment with 42× less data, with a 109M-parameter text JEPA outperforming LLMs up to 72B on cross-modal benchmarks.
CHORD提出了一种跨模型幻觉检测（CMHD）框架，将LLM隐藏状态表示为层级-词元-特征关系图，通过关系感知注意力与元学习获得可迁移到未见目标模型的检测信号，在跨模型泛化上优于现有检测方法。
The paper turns the Platonic Representation Hypothesis into a theorem, proving that JEPAs trained separately on multiple modalities converge to shared representations up to a single rotation, and shows that next-token prediction is essentially a JEPA with 'bad views' — making it far less efficient for cross-modal alignment (good views need 42x fewer samples).
The paper introduces ImagiRep, a simulator-grounded benchmark showing that spatial reasoning benefits from diverse representations beyond RGB, and proposes GEARS, a lightweight test-time strategy for adaptively selecting representations to improve reasoning about unobserved states.
提出LOTUS框架，通过低秩多线性分解从LLM隐藏状态中提取紧凑表示以检测幻觉，在多个LLM和问答基准上超越现有最先进方法。
A controlled evaluation of five public EEG foundation models plus an in-house JEPA-style model, showing that random-initialisation, band-power baselines, and validation-selected probe layers substantially alter reported pretraining gains on sleep, emotion, and motor imagery benchmarks.
该论文提出一个理论框架，分析监督对比学习目标在长尾分布下是否能形成正则单纯形几何，并提出一种完全平衡的对比目标，可证明在最优时形成正则单纯形，从而在长尾识别和迁移学习任务上取得更好的性能。
Google 推出 EmbeddingGemma 2，一个统一文本、图像、音频与视频表示的开源多模态嵌入模型，采用简化逐层嵌入与多阶段训练课程，在各模态基准上建立新的帕累托前沿，性能超越大 2 至 12 倍的模型。
This paper introduces Visible-State Supervision (ViSS) and Test-Time Causal View Search (TVS) to separate representation learning from evidence selection in streaming video understanding, achieving state-of-the-art accuracy on StreamingBench and OVO-Bench with far fewer processed frames.
OmniSpectra 提出一种波长原生的光谱基础模型，采用连续坐标与物理约束解码器实现跨仪器表征学习，仅用 12.2M 参数即在八个巡天数据集上超越更大的基线模型，表明物理结构编码可以替代参数规模。
This paper extends the Platonic Representation Hypothesis to robotics, introducing Sensorimotor Alignment (S²) as an offline proxy that measures overlap between a model's sensory representations and end-effector trajectories, showing it correlates with downstream robot task success across VLAs, WAMs, and real-robot policies.
文章介绍了一种名为"Next-Concept Modeling"的新范式，主张生成式AI应先生成抽象概念、再生成文字，而非直接以词为输出单位，有望改变现有生成式AI的建模方式。
TRACE is a training-aware benchmark framework that reorganizes 168 text-vision embedding tasks into General, Adaptability, and Challenge categories, revealing that data adaptation improves adaptability but fails to solve challenging multilingual and visual-document tasks.
论文提出正交边界遗忘（OBU）框架，通过在激活协方差的正交零空间中限制参数更新，在大模型中消除双用领域（如网络安全、生物、化学、核工程）的危险能力，同时不损害良性科学能力，避免概念混淆。
This paper classifies when the represented mobility map of parameter gradient flow is fully determined by the representation, proving an initialization classification for wide two-factor matrix representations and relating ordinary closure to invariant-level metric factorization in analytic and smooth regimes.
This paper introduces 'decision sufficiency' as a framework for evaluating whether learned representation interfaces, such as selected sparse autoencoder features, retain enough information for a controller to safely steer language models, and audits Qwen2.5, SmolLM2, and Phi showing measurable information loss in safety-critical decisions.