PulseAugur / Brief
EN
LIVE 12:05:17

Brief

last 24h
[2/2] 224 sources

Multi-source AI news clustered, deduplicated, and scored 0–100 across authority, cluster strength, headline signal, and time decay.

  1. TOOL · dev.to — LLM tag Русский(RU) ·

    Qwen in Russia: how to use and connect via API

    The Qwen family of large language models, developed by Alibaba Group, is now accessible in Russia through an API aggregator called promptra.ru. This service allows Russian users to pay for Qwen models, including the Qwen 3.6 Plus, in rubles, bypassing the need for foreign payment methods. The aggregator offers an OpenAI-compatible endpoint and mirrors Alibaba's pricing, making it a cost-effective option for developers. Qwen models are noted for their multilingual capabilities, particularly in Asian languages, strong performance in coding tasks, and a large context window of up to one million tokens. AI

    Qwen in Russia: how to use and connect via API

    IMPACT Expands access to advanced LLMs for Russian developers and businesses, potentially lowering costs and enabling new applications.

  2. RESEARCH · arXiv cs.CL Italiano(IT) · · [2 sources]

    Multimodal Evaluator Preference Collapse: Cross-Modal Contagion in Self-Evolving Agents

    A new research paper explores "Evaluator Preference Collapse" (EPC) in AI agents, finding that multimodal settings significantly amplify this bias. When using GPT-4o to evaluate DeepSeek-chat, a single strategy dominated 48.4% of the weight, a 3.2x increase compared to text-only evaluations. The study also identified "cross-modal contagion," where preferences learned in one modality transfer to and negatively impact another. Self-evaluation proved nearly immune to contagion, while cross-model evaluation was identified as the primary risk factor. AI

    IMPACT Highlights potential biases in AI systems, particularly when agents evaluate their own multimodal outputs, suggesting a need for careful design of evaluation frameworks.