Drift
PulseAugur coverage of Drift — every cluster mentioning Drift across labs, papers, and developer communities, ranked by signal.
6 day(s) with sentiment data
-
AI context engineering faces challenges with 'lost-in-the-middle' effect
Context engineering, distinct from prompt engineering, focuses on managing all inputs a model receives at inference time, including system prompts, tool definitions, and message history. A key challenge is "context rot,…
-
New DRIFT attack removes diffusion model watermarks with 98-100% success
Researchers have developed a new method called DRIFT to remove watermarks embedded in images generated by diffusion models. This black-box attack works by partially re-noising images and then using stochastic reverse re…
-
MLOps models silently drift despite dashboard uptime
The article highlights a critical issue in MLOps: while system uptime is often monitored, the performance and accuracy of deployed machine learning models can degrade silently over time. This 'model drift' can lead to s…
-
New RAG methods LiteRAG and CAGE enhance coherence and efficiency
Two new research papers, LiteRAG and CAGE, propose novel methods for improving retrieval-augmented generation (RAG) systems. LiteRAG focuses on reducing query-time costs and improving generation efficiency by using algo…
-
AI agent OAuth grants emerge as new security risk after Vercel incident
A security incident at Vercel, disclosed on April 19, 2026, highlighted a new type of indicator of compromise related to AI agent OAuth grants. The incident originated from a compromise of Context.ai, a third-party AI t…
-
Identity attackers exploit trusted vendors and forgotten credentials to breach organizations
Sophisticated attackers are increasingly targeting organizations by exploiting trusted third-party relationships and forgotten credentials, rather than directly breaching defenses. This strategy involves attackers study…
-
New DRIFT framework boosts image generation diversity and alignment
Researchers have introduced DRIFT (Diversity-Incentivized Reinforcement Fine-Tuning), a new framework designed to enhance the versatility of image generation models. This method addresses the issue of "diversity collaps…
-
Klaviyo acquires AI startup Agency to boost its AI agent capabilities
Klaviyo, an e-commerce marketing automation company, has acquired Agency, an AI-powered customer success startup. The acquisition aims to accelerate Klaviyo's development of AI agents for marketing campaigns and custome…
-
New Transformer Model DRIFT Enhances DGA Detection Against Evolving Threats
Researchers have developed a new Transformer-based framework called DRIFT to combat the evolving threat of Domain Generation Algorithms (DGAs) used in botnets. Through a nine-year study, they observed that existing DGA …
-
New research explores flow matching model enhancements and vulnerabilities · 9 sources tracked
Researchers are exploring novel approaches to enhance flow matching models, a popular paradigm for generative tasks. One paper introduces "denoising acceleration" (accel) as a cost-free proxy for estimating uncertainty …
-
New DRIFT framework enhances MRI super-resolution with adaptive flow
Researchers have developed DRIFT, a novel two-stage framework for improving the resolution of through-plane Magnetic Resonance Imaging (MRI). This method addresses the trade-off between speed and fidelity in MRI super-r…
-
New DRIFT framework boosts LLM self-improvement, sets SOTA benchmarks · 2 sources tracked
Researchers have developed DRIFT, a novel framework for enhancing large language model self-improvement without external expert supervision. DRIFT employs Difficulty Routing and Rhythm Gating to manage the model's learn…
-
MLOps: Automating Model Retraining After Drift Detection
This article discusses the importance of moving beyond simply detecting data drift in machine learning models to actively addressing it through automated retraining. It emphasizes that the ultimate goal is to ensure mod…
-
New DRIFT method refines LLM training data for improved performance
Researchers have developed DRIFT, a novel method for refining instruction data to improve the performance ceiling of large language models. Unlike existing data curation techniques that focus on subset selection, DRIFT …
-
MLOps: Beyond Model Training - A Practical Guide
Two articles discuss MLOps, focusing on the practical aspects beyond initial model training. The first article emphasizes that building an MLOps platform is a significant undertaking, with training the model being only …
-
New DRIFT method improves AI-generated image detection
Researchers have developed a new method called DRIFT for detecting AI-generated images, which adapts to unseen image generators. This approach formulates detection as learning an invariance manifold of real images using…
-
AI framework DRIFT boosts 6G satellite network efficiency
Researchers have developed a new AI-driven framework called DRIFT for predicting wireless channel responses in 6G non-terrestrial networks. This lightweight architecture aims to reduce pilot overhead by relying on data-…
-
New DRIFT framework enhances LLM multi-turn learning efficiency
Researchers have introduced DRIFT, a new framework designed to improve the efficiency of training large language models for multi-turn interactions. DRIFT addresses the trade-off between costly online reinforcement lear…