Agentic AI news, research notes, and industry analysis.
Tag
12 stories in this tag.
Figma's October 29 writeup describes how it trained its own canvas-editing model using synthetic SFT data followed by RL with designer-calibrated reward models. The results — 20% faster tasks, 60% fewer undos — are more interesting than the benchmark numbers, because undo rate is the one metric that tells you whether a model actually understood what a user meant.
Read more →A blog post from this weekend describes using GPT-6 Astra to diagnose and edit the weights of Qwen3.5-0.8B without gradient descent, raising MATH-500 accuracy from 7.9% to 48.1% by removing a single looping direction from two MLP layers. It's early and constrained, but the concept of an LLM doing targeted surgical weight editing is new territory.
Read more →The GVS5H paper showed that five Qwen3.8-27B instances coordinating through a shared file system outperform Fable 5 on LiveCodeBench Hard at roughly 9% of the cost. Memdoor, an open-source harness, independently reproduced the result — 49 of 50 runs solved, $0.39 median per problem.
Read more →Shanghai AI Lab dropped Sage-A1.5 under Apache 2.0: a 35B MoE model (3B active parameters) built on Qwen3.6 with parallel subagent orchestration and mid-task verification. It's positioned for long-horizon agentic tasks and open scientific discovery, with 256K context and 26 quantized variants available.
Read more →Alibaba shipped Qwen-Image-2.1, a technically impressive 7B image generation model with native RGBA transparency and 10-reference editing that tops the open-weight rankings — then quietly swapped the Apache 2.0 license for a research-only Qwen License. The weights are public; the use rights aren't.
Read more →A researcher prefilled Qwen3.8 A95B with the first 1% of a GPT-5.5 Pro chain-of-thought and measured how much the model continued following it. The resulting +18 pp jump in overlap — much larger than seen in other models — is a practical signal that Qwen's post-training data may include GPT-5.5 Pro reasoning traces, and points to a general method for probing where a model learned to think.
Read more →Nari Labs walked through five coordinated optimizations that bring Qwen3-TTS 1.7B to sub-50 ms p95 time-to-first-audio on a single H100, at $2 per million characters — against ElevenLabs at $100/M. None of the five changes require a new model architecture. Each targets a specific latency source, and the gains compound.
Read more →Alibaba dropped Qwen3.8-27B open weights under Apache 2.0 a day after the 2.4T Max. The 27B dense model achieves 90.3% on LiveCodeBench and 89.2% on GPQA Diamond — within touching distance of closed frontier models — while fitting on hardware that actually exists in people's garages.
Read more →PrismML released Bonsai 27B on July 14: 1-bit binary and ternary builds of Qwen3.6-27B that fit in 3.9 GB and 5.9 GB respectively, run at 11 tok/s on an iPhone 17 Pro, and retain over 90% and 95% of full-precision benchmark performance. The compression factor is around 14× versus FP16, and the models are available under Apache 2.0.
Read more →A new paper finds that reinforcement learning gains in transformers concentrate almost entirely in a narrow band of middle layers. Training just one layer at roughly 40–60% network depth can match or exceed full-parameter RL fine-tuning. The finding challenges the assumption that all layers participate equally in post-training, and has practical implications for compute-efficient alignment.
Read more →Alibaba's Qwen team released Qwen-AgentWorld, two open-weight models trained to simulate digital-agent environments — terminals, browsers, OS interfaces, software engineering tasks — via chain-of-thought reasoning. The bet is that a sufficiently accurate environment simulator lets you run RL training without real environment calls, which is expensive, slow, and hard to parallelize at scale.
Read more →Alibaba's Qwen team released Qwen-Scope, sparse autoencoder weights for Qwen3 and Qwen3.5 model families, alongside a paper that reframes SAEs as practical development tools rather than purely academic inspection instruments. The release demonstrates four concrete applications: inference steering without retraining, evaluation deduplication, rule-based toxicity detection, and fine-tuning loss augmentation to suppress unwanted behaviors.
Read more →