Data filtering works a lot worse than you would expect
This work was largely done during Neel Nanda's MATS 10.0 Exploration Phase. J Rosser and Dohun Lee are co-first authors for this post with equal contribution.…
Full feed
Feed 28,086 posts
Every story across every category, newest first. Each card links to the original publisher; daily-brief posts open as editorial pages.
When does a chess transformer “see” a knight fork? An initial result from logit lens and attention patterns
(parts 2 and 3 to follow)Summary of this postThis post is on the results of a mechanistic interpretability project aimed at understanding the internals of Maia…
OpenAI Releases GPT-Realtime-2.1 and GPT-Realtime-2.1-mini for Low-Latency Voice Agents in the API
OpenAI added two Realtime models to its API. GPT-Realtime-2.1-mini is a mini reasoning model for voice, priced like the earlier gpt-realtime-mini. OpenAI also cut p95 latency…
RT Ahmad: That’s a wrap on AIE from their new HQ! Honored to have been a guest panelist for Nvidia’s Local Al State of the Union, and to have such a…
RT AhmadThat’s a wrap on AIE from their new HQ!Honored to have been a guest panelist for Nvidia’s Local Al State of the Union, and to…
AI obsessions, continued: the act of naming something, whether something is "earned," what "carries" something else, strained metaphors, sitting with …
AI obsessions, continued: the act of naming something, whether something is "earned," what "carries" something else, strained metaphors, sitting with something (often an idea), an inanimate…
Thank you @ModelScope2022
Thank you @ModelScope2022ModelScope: Meet Hy3 on ModelScope! 295B total / 21B active MoE, built for agentic workflows with 256K context and an FP8 variant ready for…
imo this is the most impt part of anthropic's J-space paper today. it's a two-parter: 1) ant proved that they can do "brain surgery" interventions int…
imo this is the most impt part of anthropic's J-space paper today. it's a two-parter:1) ant proved that they can do "brain surgery" interventions into reasoning…
Thanks to @vllm_project for supporting Hy3! Let's play with it.
Thanks to @vllm_project for supporting Hy3! Let's play with it.vLLM: Spin it up now! 🚀
UI-MOPD: Multi-Platform On-Policy Distillation for Continual GUI Agent Learning
Recent advances in multimodal foundation models and agent systems have driven GUI agents from single-platform task execution toward cross-platform interaction. However, building multi-platform GUI agents remains…
Vision Pretraining for Dense Spatial Perception
Dense spatial perception is essential for physical intelligence, where visual systems are expected to recover structured, metric, and actionable representations from pixel observations. Modern visual foundation…
Machine Unlearning via Information Theoretic Regularization
arXiv:2502.05684v5 Announce Type: replace-cross Abstract: How can we effectively remove or ``unlearn'' undesirable information, such as specific features or the influence of individual data points, from…
Integrating Neural Encoders in Bayesian Generalized Linear Mixed Models for Multimodal Data
arXiv:2607.04647v1 Announce Type: new Abstract: Scalable Bayesian inference for generalized linear mixed models (GLMMs) provides uncertainty-aware analysis of correlated longitudinal data, but existing scalable approaches largely…
Causal ASCEND: Scalable Two-tier Causal Discovery on High Dimensional Multi-omics Data
arXiv:2607.04527v1 Announce Type: new Abstract: Biological systems exhibit a hierarchical structure, characterised by directed flow from upstream regulators to downstream effects. Although this ordering provides a…
Tightening the Score Matching Gap for Diffusion Models
arXiv:2607.04442v1 Announce Type: new Abstract: Diffusion models (DMs) are a state-of-the-art generative method to approximately sample from an unknown distribution. Their training and evaluation primarily rely…
MetaCaDI: A Meta-Learning Framework for Causal Discovery from Multiple Environments with Unknown Interventions
arXiv:2510.22298v2 Announce Type: replace Abstract: Uncovering the causal mechanisms of complex real-world systems remains a significant challenge, as these systems often entail high data collection costs…
Discrete distributions are learnable from metastable samples
arXiv:2410.13800v4 Announce Type: replace Abstract: Physically motivated stochastic dynamics are widely used to sample from high-dimensional distributions. However, such samplers often get trapped in metastable states,…
Optimal Mixture-of-Experts Model Averaging for Conditional Generative Models
arXiv:2607.04360v1 Announce Type: new Abstract: Conditional generative models have emerged as powerful tools for sampling from target conditional distributions, driving substantial advances across a wide range…
Distribution-free Deviation Bounds and The Role of Domain Knowledge in Learning via Model Selection with Cross-validation Risk Estimation
arXiv:2303.08777v3 Announce Type: replace Abstract: Cross-validation is one of the most widely used tools for risk estimation and model selection in statistics and machine learning, yet…
Contaminated Multi-task Learning with Heterogeneity: Fundamental Limits and Optimal Algorithms
arXiv:2607.02681v1 Announce Type: new Abstract: Integrating information across related tasks can improve estimation and prediction in transfer, multi-task, and federated learning, but contamination and heterogeneity make…
Denoised Conformal Alignment for Reliable Selection of Conditional Average Treatment Effect Predictions
arXiv:2607.03161v1 Announce Type: new Abstract: In selective deployment, practitioners act only on a model-chosen subset of individuals based on predicted conditional average treatment effects, but marginal…
A Hierarchy of Policy Learning Problems
arXiv:2607.03385v1 Announce Type: new Abstract: Policy learning has received substantial attention with the goal of learning policies from observational data for decision-making. A majority of work…
Sequential Correlations Change In-Context Learning: Effective Context Length and Architectural Mismatch
arXiv:2607.03660v1 Announce Type: new Abstract: Modern sequence models have a striking capacity for in-context learning (ICL); they can perform new tasks based only on examples given…
Missing Data Imputation under Manifold Hypothesis
arXiv:2607.03641v1 Announce Type: new Abstract: The manifold hypothesis posits that high-dimensional data are concentrated near a low-dimensional embedded manifold. Recent advances in mixture variational autoencoders (VAEs)…
Fixed-Confidence Best-Arm Identification for Causal Mediation Analysis
arXiv:2607.04315v1 Announce Type: new Abstract: This paper studies the problem of identifying the treatment that maximizes the expected natural direct potential outcome (NDPO), which captures the…