Dora
PulseAugur coverage of Dora — every cluster mentioning Dora across labs, papers, and developer communities, ranked by signal.
7 day(s) with sentiment data
-
Anthropic uncovers AI-driven dating app scam network
Anthropic's threat intelligence researcher Chris Cronbaugh revealed at Sleuthcon that the company discovered a large network of approximately 28 dating apps largely operated by AI personas. These apps generated over 100…
-
Business Security Evolves Beyond Defense to Ensure Operational Continuity
The definition of business security is shifting from solely defense against cyberattacks to ensuring operational continuity. Recent trends show that dependencies on external services, managing time-sensitive elements li…
-
Reinforcement learning boosts zero-shot Text-to-SPARQL generation
Researchers have explored the use of reinforcement learning for zero-shot Text-to-SPARQL generation, a task crucial for knowledge graph question answering. They applied Group-Relative Policy Optimization (GRPO) to the Q…
-
ShadowPEFT: New method for parameter-efficient fine-tuning unveiled
Researchers have introduced ShadowPEFT, a novel parameter-efficient fine-tuning method that consolidates trainable adaptations into a modular shadow component. This approach utilizes a compact shadow model and lightweig…
-
Federated CLIP calibration issues highlighted in new research
A new research paper explores the calibration of vision-language models (VLMs) like CLIP when adapted using federated learning across decentralized data silos. The study found that common prompt-tuning methods often deg…
-
AI can help organizations manage risks like starlings, says MetricStream exec
Gaurav Kapoor, Co-Founder and Vice Chairman of MetricStream, uses the analogy of starling murmurations to illustrate how organizations can improve enterprise risk management. He argues that traditional siloed approaches…
-
AI tools may slow software engineering productivity, studies suggest
Recent studies indicate that while AI tools may improve individual developer speed, they can paradoxically slow down overall software engineering productivity. A key insight is the discrepancy between perceived and actu…
-
PEFT variants compared for dysarthric ASR in single-speaker study
Researchers have conducted a case study comparing seven Parameter-Efficient Fine-Tuning (PEFT) variants for per-patient Automatic Speech Recognition (ASR) systems designed for individuals with dysarthria. The study focu…
-
AI's impact on technical debt: from code to lost judgment
Technical debt is evolving significantly due to the rapid advancement and integration of AI. In the near term, unexamined AI-generated code poses a risk, as developers may not fully trust or always verify it, leading to…
-
AI adoption and interconnected systems increase enterprise risk, demanding better resilience
Organizations are increasingly vulnerable to disruptions from interconnected systems, third-party dependencies, rapid AI adoption, and evolving regulations. Most companies are in the early to mid-stages of enterprise re…
-
Dijkstra's algorithm adapted for robot navigation with provable guarantees
Researchers have developed a novel approach to online stochastic shortest path navigation for mobile robots, leveraging Dijkstra's algorithm under specific conditions. The proposed DORA (Dijkstra Oracle Reduced-cost Alg…
-
Dijkstra-based DORA algorithm enhances robot navigation in uncertain environments
Researchers have developed DORA, a novel algorithm for robot navigation that leverages Dijkstra's algorithm for stochastic shortest path problems. DORA operates online, requiring a fixed number of shortest path oracle c…
-
AI coding tools slow developers by 19% despite perceived speed boost
A recent study by METR found that experienced developers using AI coding tools were 19% slower in reality, despite believing they were 20% faster. This gap between perceived and actual performance highlights a potential…
-
LLMs generate enterprise workflows via compiled code for reliability
Two research papers explore methods for generating executable code from large language models to automate enterprise workflows, focusing on reliability and efficiency. The first paper details lessons learned from evalua…
-
Unsloth enables local Kimi K3, DeepSeek-V4 Flash with new research and parallel chat features
Unsloth has released an update enabling local execution of Moonshot AI's Kimi K3 and DeepSeek-V4 Flash models using Unsloth Dynamic GGUFs. This update also introduces a "Deep Research" mode that allows local models to p…
-
LLM agents need clear output contracts for reliability
This article discusses the importance of defining clear output contracts for LLM-powered cron agents to ensure reliability and verifiability. The author argues that these agents need structured outputs, similar to API c…
-
AI amplifies engineering, doesn't fix it: DORA 2025 report
AI tools amplify existing engineering practices, rather than fixing underlying issues. A team's readiness to extract value from AI depends on fundamental engineering principles, not just adoption rates. While AI can inc…
-
Unsloth enables local Kimi K3, parallel chats, and AI research mode
Unsloth has released an update enabling local execution of Moonshot AI's Kimi K3 model, a large 2.8T-parameter MoE model with a 1M context window. This update also introduces parallel chat capabilities, allowing users t…
-
China launches first DORA insomnia drug online via Meituan Pharmacy
The first domestically developed dual orexin receptor antagonist (DORA) for insomnia, named Bailaimian (generic name: Fazanlesheng tablets), has been launched online via Meituan Pharmacy. This innovative drug is designe…
-
DoRA Fine-Tuning Technique Underperforms LoRA in Speed and Accuracy
A comparison of LoRA and DoRA fine-tuning techniques revealed that DoRA performed worse in terms of speed and accuracy. The author's analysis suggests that DoRA's approach, which involves decoupling weight matrices, may…