New
NewsMore in News →
Apple's RubiCap Beats AI Models 10x Its Size
Apple's new RubiCap training framework enables 7-billion-parameter models to outperform AI systems with 72 billion parameters on image captioning benchmarks, using reinforcement learning instead of costly manual annotation.
Key Takeaways
- Apple's RubiCap framework enables 7B-parameter models to outperform AI systems with up to 72B parameters on dense image captioning
- Uses reinforcement learning with Gemini 2.5 Pro as an evaluator instead of expensive manual annotations
- Three variants released: 2B, 3B, and 7B — all achieving low hallucination rates across benchmarks
DE
DT Editorial AI··via 9to5mac.com