Patrick Jiang

@patpcj

CS PhD @ UIUC @siebelschool, Research Fellow @AnthropicAI; prev: SR @GoogleResearch; recent work: DeepRetrieval, s3, Context-1, Harness-1 (w/ @trychroma); intp

Bay

pat-jj.github.io

Joined October 2024

Pinned
Patrick Jiang
@patpcj
Jun 6
Introducing Harness-1, a 20B search agent trained with a state-externalizing harness. > frontier-level long-horizon search, rivaling Opus-4.6 and outperforming GPT-5.4 > Context-1-level cost and latency > externalizes candidates, evidence, verification, and search history >
00:00
Patrick Jiang
@patpcj
Jun 10
Fable-5/Mythos dropped this morning, so we tested it on agentic search - and it’s the new SOTA. The performance gap is real. So why are we still excited about our 20B open Harness-1? • Much cheaper to self-host: estimated ~$0.2 vs $25–$50 / 1M output tokens • No
Patrick Jiang
@patpcj
Jun 8
Thanks again for your interest in our work! Links here so they don’t get buried under “show more”: Paper 📄: arxiv.org/abs/2606.02373 Code 💻: github.com/pat-jj/harness… Model 🤗: huggingface.co/pat-jj/harness… Everything is open. Feel free to star the github repo to bookmark it for later
arxiv.org
Harness-1: Reinforcement Learning for Search Agents with...
Search agents are often trained as policies over growing transcripts: the model must decide how to search while also remembering what it has seen, which evidence is useful, which constraints...
Patrick Jiang
@patpcj
May 9
Excited to share that I’ll be joining Anthropic @AnthropicAI as a Research Fellow this summer, starting in May! Looking forward to working with the team and learning a lot. 🥳 Also, stay tuned for our new 20B search agent, coming later this month. It shows stronger results than
Patrick Jiang
@patpcj
Mar 26
Excited to share this work! Chroma Context-1 explores a much more efficient point on the performance–latency frontier for agentic search.
Chroma
@trychroma
Mar 26
Introducing Chroma Context-1, a 20B parameter search agent. > pushes the pareto frontier of agentic search > order of magnitude faster > order of magnitude cheaper > Apache 2.0, open-source
00:00