February 2026

:::info Authors: Yue Wang, wang.y@salesforce.com (Salesforce Research Asia) Weishi Wang, weishi.wang@salesforce.com (Salesforce Research Asia; Nanyang Technological University, Singapore) Shafiq Joty, sjoty@salesforce.com (Salesforce Research Asia; Nanyang Technological University, Singapore) Steven C.H. Hoi, shoi@salesforce.com (Salesforce Research Asia) ::: Abstract Pre-trained models for Natural Languages (NL) like BERT and GPT have been recently shown to transfer well to Programming Languages (PL) and largely benefit a broad set of code-related tasks. Despite their success, most current methods either rely on an encoder-only […]

Ver mais

:::info Authors: Liang Wang (Microsoft Corporation) Nan Yang (Microsoft Corporation) Xiaolong Huang (Microsoft Corporation) Binxing Jiao (Microsoft Corporation) Linjun Yang (Microsoft Corporation) Daxin Jiang (Microsoft Corporation) Rangan Majumder (Microsoft Corporation) Furu Wei (Microsoft Corporation) ::: Abstract This paper presents E5 1, a family of state-of-the-art text embeddings that transfer well to a wide range of tasks. The model is trained in a contrastive manner with weak supervision signals from our curated large-scale text pair dataset (called CCPairs). E5 […]

Ver mais

TL;DR: We ran a factorial study on policy conditioning (appending a “goal” signal to observations). We found that while it barely improves “tracking precision,” it leads to a 23x improvement in tail-risk (CVaR). Crucially, we prove that temporal correlation—not just having the extra data—is the causal driver. The Problem: The “Black Box” of Conditioning In RL, we often append a task descriptor (goal, context vector, or latent) to the agent’s observation. We assume it helps the agent adapt. […]

Ver mais

:::info Authors: Chengxuan Ying, yingchengsyuan@gmail.com (Dalian University of Technology) Tianle Cai, tianle.cai@princeton.edu (Princeton University) Shengjie Luo, luosj@stu.pku.edu.cn (Peking University) Shuxin Zheng, shuz@microsoft.com (Microsoft Research Asia) Guolin Ke, guoke@microsoft.com (Microsoft Research Asia) Di He, dihe@microsoft.com (Microsoft Research Asia) Yanming Shen, shen@dlut.edu.cn (Dalian University of Technology) Tie-Yan Liu, tyliu@microsoft.com (Microsoft Research Asia) ::: Abstract The Transformer architecture has become a dominant choice in many domains, such as natural language processing and computer vision. Yet, it has not achieved competitive performance […]

Ver mais

Every crypto bull run created a new wave of millionaires. But those millionaires did not buy after the rally started. They bought during the fear. They bought presales at fractions of a cent while everyone else waited for confirmation. By the time confirmation arrived, the 50x windows had already closed. A presale is when a project sells tokens before they list on exchanges. The price is locked far below what it trades for on launch day. Early SHIB […]

Ver mais

Few-shot text classification aims to recognize unseen classes with limited labeled text samples. Existing approaches focus on boosting meta-learners by developing complex algorithms in the training stage. However, the labeled samples are randomly selected during the testing stage, so they may not provide effective supervision signals, leading to misclassification. To address this issue, we propose a textbf{L}abel-guided textbf{D}istance textbf{S}caling (LDS) strategy. The core of our method is exploiting label semantics as supervision signals in both the training and […]

Ver mais

Semantic IDs (SIDs) are compact discrete representations derived from multimodal item features, serving as a unified abstraction for ID-based and generative recommendation. However, learning high-quality SIDs remains challenging due to two issues. (1) Collision problem: the quantized token space is prone to collisions, in which semantically distinct items are assigned identical or overly similar SID compositions, resulting in semantic entanglement. (2) Collision-signal heterogeneity: collisions are not uniformly harmful. Some reflect genuine conflicts between semantically unrelated items, while others […]

Ver mais

Catastrophic forgetting, the tendency of neural networks to forget previously learned knowledge when learning new tasks, has been a major challenge in continual learning (CL). To tackle this challenge, CL methods have been proposed and shown to reduce forgetting. Furthermore, CL models deployed in mission-critical settings can benefit from uncertainty awareness by calibrating their predictions to reliably assess their confidences. However, existing uncertainty-aware continual learning methods suffer from high computational overhead and incompatibility with mainstream replay methods. To […]

Ver mais

Catastrophic forgetting, the tendency of neural networks to forget previously learned knowledge when learning new tasks, has been a major challenge in continual learning (CL). To tackle this challenge, CL methods have been proposed and shown to reduce forgetting. Furthermore, CL models deployed in mission-critical settings can benefit from uncertainty awareness by calibrating their predictions to reliably assess their confidences. However, existing uncertainty-aware continual learning methods suffer from high computational overhead and incompatibility with mainstream replay methods. To […]

Ver mais

Photo by Resource Database on Unsplash Authors: Shon Mohsin, D. Eng AIJeremiah Lowhorn, D. Eng AISeth ThorMatthew Morais, D. Eng AI What if you could ask your data: “What did we know about this research topic in June 2024?” — and get a precise, structured answer? Scientific Research generates a web of interconnected entities — subjects, methods, outcomes, organizations, publications — that evolves over time. A spreadsheet can track individual facts. A relational database can join tables. But neither preserves the shape of knowledge: the web […]

Ver mais