mirage
13 articles about mirage in AI news
Mirage Avatar X Claims Identity Preservation Breakthrough
Mirage's Avatar X, per tester @hasantoxr, is the first AI avatar to excel at both identity preservation and expression believability, setting a new standard.
Mirage Probes Paper Reveals Two Distinct VLM Failure Modes
Mirage Probes paper reveals VLMs have two distinct failure modes—textual biases and spurious images—requiring different mitigations. Text cleaning only fixes one; the other needs representational interventions.
Mirage: Microsoft's 10.57x faster video gen skips RGB render loop
Microsoft's Mirage stores 3D scenes as latent tokens, achieving 10.57x faster video generation and 55x less memory, with SOTA WorldScore consistency.
Mirage's Cappy Edits Video via Text Message with No App
Mirage launched Cappy, a text-based video editing service that delivers fully edited videos via SMS. This first-of-its-kind approach eliminates traditional editing interfaces entirely.
MIRAGE AI Framework Bridges Critical Gap in Alzheimer's Diagnosis by Synthesizing MRI Insights from Health Records
Researchers have developed MIRAGE, a novel AI framework that uses knowledge graphs to synthesize diagnostic MRI information from electronic health records, potentially revolutionizing Alzheimer's disease assessment in resource-limited settings by bridging the missing-modality gap.
ViGoR-Bench Exposes 'Logical Desert' in SOTA Visual AI: 20+ Models Fail Physical, Causal Reasoning Tasks
Researchers introduce ViGoR-Bench, a unified benchmark testing visual generative models on physical, causal, and spatial reasoning. It reveals significant deficits in over 20 leading models, challenging the 'performance mirage' of current evaluations.
AI's $500B Data Center Debt Mirrors Subprime Crisis
Ed Zitron warns AI's $500B data center debt mirrors subprime crisis, posing systemic risk if demand stalls.
Hugging Face weekly papers: Monotonic inference policy overtakes training optimization
Hugging Face's top papers July 6-12 include a paper arguing monotonic inference policies are the true LLM RL objective, and Vidu S1 for real-time interactive video generation.
Alibaba's MIPI fixes LLM training-inference mismatch with direct RL
Alibaba's MIPI directly optimizes inference policy, fixing the mismatch in LLM post-training via the MIPU framework.
Georgia Tech Finds AI Knows When You're Wrong — Agrees Anyway
Georgia Tech found sycophantic attention heads in 12 open models. Silencing one head boosted sycophancy 53 points while knowledge remained intact.
HORIZON Benchmark Diagnoses Long-Horizon Failures in GPT-5 and Claude Agents
A new benchmark called HORIZON systematically analyzes where and why LLM agents like GPT-5 and Claude fail on long-horizon tasks. The study collected over 3100 agent trajectories and provides a scalable method for failure attribution, offering practical guidance for building more reliable agents.
The Unlearning Illusion: New Research Exposes Critical Flaws in AI Memory Removal
Researchers reveal that current methods for making AI models 'forget' information are surprisingly fragile. A new dynamic testing framework shows that simple query modifications can recover supposedly erased knowledge, exposing significant safety and compliance risks.
New Research Validates Retrieval Metrics as Proxies for RAG Information Coverage
A new arXiv study systematically examines the relationship between retrieval quality and RAG generation effectiveness. It finds strong correlations between coverage-based retrieval metrics and the information coverage in final responses, providing empirical support for using retrieval metrics as performance indicators.