BLEU, or Bilingual Evaluation Understudy, is an automatic metric originally developed for machine translation. It compares short sequences of words in generated text with one or more reference translations, combines precision across several n-gram lengths, and applies a penalty when output is too short. BLEU is inexpensive and useful for comparing systems on the same dataset, but it does not directly measure factuality, fluency, meaning, or human preference. Valid alternative wording can receive a low score, while awkward text can match many reference phrases. Results depend on tokenization and implementation, so reporting should include the exact evaluation settings and, when possible, human judgment.
Qualcomm agreed to acquire Modular, the AI software startup led by Chris Lattner, to build a hardware-agnostic software layer and challenge Nvidia in data center AI.
ByteDance’s record offshore loan would nearly double its 2024 facility and help fund capital spending that could reach $70 billion this year on AI data centers.
ByteDance unveiled Doubao 2.1 Pro, which it claims beats Claude Opus 4.6 at 80% lower cost, and previewed Seedance 2.5, a 30-second video model.
A new Anthropic privacy policy taking effect July 8 lets the company ask some flagged Claude users to upload government IDs and submit biometric selfies.
The back-to-back exits of Noam Shazeer and AlphaFold’s John Jumper expose Google’s retention struggles as funded rivals lure away frontier AI researchers.
Google is investing about $75 million in indie studio A24, pairing it with DeepMind to build AI filmmaking tools in its first stake in a studio.
OpenAI added Record and Replay to its Codex Mac app, letting users demonstrate a task once and turn it into a reusable AI skill.
Salesforce agreed to buy Fin, the customer-service AI company formerly known as Intercom, for about $3.6 billion to strengthen its Agentforce platform.
Kazakhstan’s Data Center Valley would deploy 100,000 Nvidia GPUs near cheap power by 2027, though analysts flag execution risk after a history of stalled deals
Z.ai released GLM-5.2, a flagship coding model with a usable 1-million-token context, shipping it to paying subscribers first while withholding benchmarks, an API and open weights.