#qwen3
Qwen3: 6 AI articles covering qwen3 news, analysis, and research
Articles
Stanford and UC Berkeley Researchers Unveil TokenGate: Tokenβ8
Stanford and UC Berkeley researchers unveil TokenGate, a token-gated attention module that accelerates long-context LLM prefill up to 11.3x with no accuracy los...
NVIDIA PivotOPD Teaches Multi-Turn AI Agents to Recover Fromβ8
NVIDIA's PivotOPD trains multi-turn AI agents to avoid pivotal mistakes and recover from them, beating 13 baselines on ALFWorld, WebShop, and Search QA.
The Story of Qwen: Alibaba's AI Models From 7B to 2.4Tβ7
From a 7B chatbot in 2023 to open-weight models with 2.4 trillion parameters, here's the story of how Alibaba's Qwen family evolved release by release.
The First Token Is Not the Verdict: Hidden Costs of Reading LLMβ8
Reading LLM judge verdicts from the first token overstates position bias by up to 42 points, per new researchβa hidden cost for anyone auditing judge reliabilit...
H Company Unveils Holo4: Open-Weight Computer-Use Models Thatβ9
H Company's Holo4 open-weight computer-use models bring desktop, web, Android, and API automation to one Apache 2.0 agent stack with 256K context and strong OSW...
Fine-Tuning Tool-Calling LLMs: A Complete Guide Using XYZ-Aquilaβ10
Fine-tune Qwen3-0.6B with LoRA on XYZ-Aquila-SFT tool-calling data. Full pipeline from preprocessing to evaluation for skill enhancement.
