Haoming Jiang
@jiang_haoming
Feeding AGI at OpenAI | Posttraining Research
We're excited to announce we’ve launched several improvements to ChatGPT search, and today we’re starting to roll out a better shopping experience. Search has become one of our most popular & fastest growing features, with over 1 billion web searches just in the past week 🧵
Ever notice it’s hard to get great answers on shopping journey questions? And, to get good answers where they’re deeply integrated into a shopping experience you love and trust? Buying guidance like: “what is important to consider when buying headphones?” Shopping missions…
I asked ChatGPT agent to train a machine learning model and asked it to improve the model! AI trains and improves AI — AGI is coming! @OpenAI
ChatGPT can now do work for you using its own computer. Introducing ChatGPT agent—a unified agentic system combining Operator’s action-taking remote browser, deep research’s web synthesis, and ChatGPT’s conversational strengths.
First attempt at fixing the MCP issue using Codex, feeling that AGI moment again! github.com/modelcontextpr…
Today we're unveiling OLMoTrace, a tool that enables everyone to understand the outputs of LLMs by connecting to their training data. We do this on unprecedented scale and in real time: finding matching text between model outputs and 4 trillion training tokens within seconds. ✨
For years it’s been an open question — how much is a language model learning and synthesizing information, and how much is it just memorizing and reciting? Introducing OLMoTrace, a new feature in the Ai2 Playground that begins to shed some light. 🔦
I'll be attending the amazing CONFERENCE ON LANGUAGE MODELING. Let me know if you're planning to attend so that we can say Hi! 👋 Or register now and join me at the event! whova.com/portal/registr… via #Whova event app
Following MemoryLLM (arxiv.org/abs/2402.04624), We have trained a new model memoryllm-8b (huggingface.co/YuWangX/memory…) based on Llama3 with a memory pool of size 1.67B! Based on this we built a Chat model memoryllm-8b-chat(huggingface.co/YuWangX/memory…) Check them out!
The inductive and deductive capability of a language model might just be duality principle of how we should think of reasoning capability. arxiv.org/pdf/2408.00114 forbes.com/sites/lanceeli…
🥸
WHAT? @inflectionAI is just a claude-3-sonnet wrapper? care to explain? 🐒 Produces the exact same answer word to word for a custom query i asked 🤯
Despite the mixed feelings about Google's latest Gemma model, we're big fans! @GoogleAI Why? Coz we found it pairs incredibly well with our SelfExtend 🤣🤣🤣 - like, perfectly! With Self-Extend, no fine-tuning needed, we effortlessly expanded Gemma's window from 8k to 90k+! On…
Announcing Stable Diffusion 3, our most capable text-to-image model, utilizing a diffusion transformer architecture for greatly improved performance in multi-subject prompts, image quality, and spelling abilities. Today, we are opening the waitlist for early preview. This phase…
In crafting LLM alignment data, striking the right balance is key. We should think some minimum sufficient criteria. Extra in responses = Hallucinations 🚫. Extra in instructions = Spurious biases 🔄.