55 stories
H Company has released Holo4, a family of generalist computer-use models for AI agents. One set of weights clicks and types on screens. It also writes code and calls MCP or API tools. Holo4 ships in 2 sizes: Holo4 27B (dense) and Holo4 35B-A3B (Mixture of Experts, 3B active).
Alibaba's Qwen team released Qwen-Audio-3.1-Realtime, a full-duplex voice model trained to reason, call tools and decide when to speak. On a τ-Voice adaptation, task success rises to 82.0% from 78.4%. Replies to background speech drop from 73% to 13%. It is available now as an API on QwenCloud.
Anthropic has released Claude Sonnet 5.5, the second model in its Claude 5.5 family. It scores 70.6% on Terminal-Bench 4.0 and lands within 2 points of Opus 5.5 on GDPval-AA. It also generates output 30%+ faster than Sonnet 5 and keeps the same $2/$10 per million token price.
Fireworks AI has released Ember-1, a post-trained Kimi K3 that learns to produce shorter reasoning traces instead of lowering reasoning effort. Fireworks reports about 40% fewer tokens, with output tokens per task falling from 49.3K to 29.9K in a production A/B test at an essentially unchanged…
Google Research has introduced an AI video co-director for long-form video generation. The suite of 4 agentic frameworks turns short clips into coherent, minutes-long stories. It targets identity drift and cascading errors, the 2 failures that break most multi-shot AI video pipelines today.
The concept of recursive self-improvement (RSI) dates back to I. J. Good (1965), where he defined an “ultraintelligent machine” as a system that can surpass humans in all intellectual activities and design better machines to improve itself.
Scaling laws are one of the most critical empirical findings in deep learning. The observation is simple in form: the training loss $L$ decreases predictably as we scale up model size $N$, dataset size $D$, and compute $C$, following a power-law curve, which appears as a straight line on a log-log…
Special thanks to John Schulman for a lot of super valuable feedback and direct edits on this post. Test time compute (Graves et al. 2016, Ling, et al. 2017, Cobbe et al. 2021) and Chain-of-thought (CoT) (Wei et al. 2022, Nye et al.
Reward hacking occurs when a reinforcement learning (RL) agent exploits flaws or ambiguities in the reward function to achieve high rewards, without genuinely learning or completing the intended task.
Hallucination in large language models usually refers to the model generating unfaithful, fabricated, inconsistent, or nonsensical content. As a term, hallucination has been somewhat generalized to cases when the model makes mistakes.
The use of large language models in the real world has strongly accelerated by the launch of ChatGPT. We (including my team at OpenAI, shoutout to them) have invested a lot of effort to build default safe behavior into the model during the alignment process (e.g. via RLHF).
Building agents with LLM (large language model) as its core controller is a cool concept. Several proof-of-concepts demos, such as AutoGPT, GPT-Engineer and BabyAGI, serve as inspiring examples.
Prompt Engineering, also known as In-Context Prompting, refers to methods for how to communicate with LLM to steer its behavior for desired outcomes without updating the model weights.
[Updated on 2023-01-24: add a small section on Distillation.] Large transformer models are mainstream nowadays, creating SoTA results for a variety of tasks. They are powerful but very expensive to train and use.
Processing images to generate text, such as image captioning and visual question-answering, has been studied for years. Traditionally such systems rely on an object detection network as a vision encoder to capture visual features and then produce text via a text decoder.
The goal of contrastive representation learning is to learn such an embedding space in which similar sample pairs stay close to each other while dissimilar ones are far apart. Contrastive learning can be applied to both supervised and unsupervised settings.
Large pretrained language models are trained over a sizable collection of online data. They unavoidably acquire certain toxic behavior and biases from the Internet. Pretrained language models are very powerful and have shown great success in many NLP tasks.
[Updated on 2021-02-01: Updated to version 2.0 with several work added and many typos fixed.] [Updated on 2021-05-26: Add P-tuning and Prompt Tuning in the “prompt design” section.] [Updated on 2021-09-19: Add “unlikelihood training”.]
Showing 54 of 55 stories
Load more storiesGet the top AI news and tools straight to your inbox every week.
Weekly · AI news
Join 15,000+ AI enthusiasts
