Tencent releases EVIE-8B and EVIE-4.5B for visual document retrieval
Tencent ships EVIE-8B and EVIE-4.5B visual document retrieval models, with EVIE-8B hitting 66.75 nDCG@10 on ViDoRe V3.
Read digestAI News Feed
Page 16 of 19
Tencent ships EVIE-8B and EVIE-4.5B visual document retrieval models, with EVIE-8B hitting 66.75 nDCG@10 on ViDoRe V3.
Read digestSony Music Publishing and Warner Chappell sued Anthropic for alleged copyright infringement in training Claude with copyrighted songs.
Read digestA radar-only classification study shows point density, not MLP architecture, is the key driver of accuracy gains.
Read digestQwen3.8 Flash Next hits 45 tok/s on an M4 Max, making the 125B MoE model practical for local agentic workloads.
In this digest
Two more publishers sue OpenAI and Microsoft over copyrighted journalism, while hikers stranded on Mount Shasta blame Gemini's route advice.
In this digest
OpenAI revealed an autonomous-agent wiki incident and is creating a new framework for AI misalignment disclosures.
In this digest
Nscale targets up to $3.5B ahead of a U.S. listing, while a Reddit user benchmarks 21 Qwen3.8-27B quants on 16GB VRAM.
In this digest
Tencent released Hy4 preview, a massive 770B-parameter MoE model with a 1M-token context window and open weights under Apache 2.0 license.
In this digest
Ugreen launches local-AI smart-home hubs, and TrueSpar open-sources Paddock, a Rust/C++ inference engine beating vLLM in benchmarks.
In this digest
Thinking Machines seeks $1B at a $40B valuation, a runtime MoE tweak cuts Qwen reasoning tokens, and ampixa ships a tiny multilingual TTS stack.
In this digest
Put practical data boundaries and usage policies in place before AI makes existing oversharing easier to find.
Explore Secure AI Adoption