Apple to present 60 AI research studies at ICLR 2026, including SHARP 3D reconstruction model
Apple will present nearly 60 research studies and technical demonstrations at the International Conference on Learning Representations (ICLR) running April 23-27 in Rio de Janeiro. Demos include the SHARP model that reconstructs photorealistic 3D scenes from a single image in under one second, running on iPad Pro with M5 chip.
Apple to present 60 AI research studies at ICLR 2026, including SHARP 3D reconstruction model
Apple will present nearly 60 research studies and technical demonstrations at the International Conference on Learning Representations (ICLR) running April 23-27 in Rio de Janeiro, Brazil.
Key demonstrations
The company's technical demos at booth #204 will showcase:
- SHARP model: Reconstructs photorealistic 3D scenes from a single image in under one second, running on iPad Pro with M5 chip
- On-device LLM inference: Demonstration of a quantized frontier coding model running entirely locally on MacBook Pro with M5 Max using MLX, Apple's open-source framework for AI inference on Apple silicon, within Xcode's native development environment
The SHARP model uses depth estimation technology to generate 3D Gaussian representations from single images, according to a December 2025 paper from Apple researchers.
Conference participation
Apple is serving as one of the sponsors for ICLR 2026. The company's participation includes:
- Poster presentations
- Oral presentations
- Workshop sessions
- Technical demonstrations during exhibition hours
The presentations span research across machine learning and AI, with many studies previously covered by technology media outlets.
MLX framework
Apple's MLX is an open-source framework built specifically for AI inference on Apple silicon. The framework enables developers to run large language models and other AI workloads locally on Mac hardware without cloud dependencies.
The technical demonstration will show a quantized coding model running within Xcode, Apple's integrated development environment for building applications.
What this means
Apple's substantial research presence at ICLR 2026 signals continued investment in AI capabilities, particularly in on-device inference and efficient model deployment. The SHARP model's sub-second 3D reconstruction demonstrates progress in computationally intensive tasks running on mobile hardware. The company's focus on local inference using MLX aligns with its privacy-oriented approach, differentiating from cloud-dependent AI strategies of competitors. With 60 studies presented, Apple is positioning itself as a significant research contributor in the academic AI community.
Related Articles
Apple Releases LensVLM-9B, a 9B Vision-Language Model That Selectively Decompresses Text Images
Apple has released LensVLM-9B, a 9-billion-parameter vision-language model fine-tuned from Qwen3.5-9B-Base that processes documents as compressed images, selectively expanding only relevant pages to full resolution. The model supports 5x, 10x, and 15x compression ratios and is available under Apple's Machine Learning Research Model License.
Goodfire researcher claims 90% of Qwen-32B neurons have one job, challenging polysemanticity
Nick Cammarata of Goodfire published an essay, "Neural Networks Are Already Clean," arguing that agentic analysis resolves many known cases of polysemanticity and superposition. A widely shared summary says 90% of neurons in Qwen-32B have a single function. The results are unreplicated, and key methodology was not available in the source material.
Apple brings Siri AI beta to Apple TV 4K 2nd and 3rd generation, alongside new $299 A19 model
Apple says Siri AI will be available as a beta on Apple TV 4K (2nd generation or later) through a forthcoming tvOS update. The company also introduced a fourth-generation Apple TV 4K with an A19 chip and a $299 starting price.
Sakana AI's MASS self-improvement lifts Qwen3.6-27B win rate to 69.9%, but self-judging passes flawed runs
Researchers from Sakana AI and UC Berkeley introduced Multi-Agent Self-Supervision (MASS), in which a single model solves tasks, judges results and optimizes workflows. Applied to Qwen3.6-27B, the win rate against the base model reached 69.9% after two rounds. But the self-evaluator sometimes accepted flawed or manipulated solutions as successes and fed them into the next training cycle.
Comments
Loading...