Apple releases videos from its 2025 AI Reasoning and Planning Workshop
Source: 9to5Mac
The newly published videos from Apple’s two‑day event focus on three key areas related to AI: Reasoning and Planning, Applications to Agents, and Model Development. Here are the details.
Valuable material, despite being seven months old
Last July, Apple hosted an event called the Workshop on Reasoning and Planning, featuring presentations and discussions from Apple engineers and researchers from UCLA, Stanford University, UC Berkeley, among others.
The event featured nearly 30 studies, which were presented and discussed over the course of two days, all of which can be found here.
Apple also shared eight talks from the event, featuring the following:
- From LLMs to Embodied AI Agents: Lessons and Methods, presented by Alexander Toshev (Apple)
- MMAU: A Holistic Benchmark of Agent Capabilities Across Diverse Domains, presented by Yanchao Sun (Apple)
- More Efficient and Accurate Reasoning with Adaptive Parallel Reasoning, presented by Alane Suhr (UC Berkeley)
- On Evaluating and Improving Robustness in Visual Reasoning Abilities of Vision‑Language Models, presented by Melanie Mitchell (Santa Fe Institute)
- Open‑Ended and AI‑Generating Algorithms in the Era of Foundation Models, presented by Jeff Clune (University of British Columbia)
- Reasoning, Intelligence & Large Language Models, presented by Iman Mirzadeh (Apple)
- Reinforcement Learning for Long‑Horizon Interactive LLM Agents, presented by Philipp Krähenbühl (Apple)
- Towards Internet‑Scale Training For Agents, presented by Ruslan Salakhutdinov (CMU)
Despite the seven‑month delay between these talks and today’s publication of the videos (which feels long given how fast these technologies are advancing), they are still worth watching if you’re interested in AI reasoning, planning, and agentic systems.
They offer valuable lessons on how Apple’s engineers and academic researchers are thinking about the state of the art in these areas, as well as the open problems shaping where these technologies may go next.