Pair-free 109M text-JEPA hits 64% zero-shot on ImageNet100, beating 8B LLMs
randall_balestr · x · 2026-10-30
Additional result from the thread: on the latest pair-free benchmark from Phillip Isola's group, the 109M text-JEPA achieves 64% accuracy on ImageNet100, beating every zero-shot LLM result reported there (up to 8B, 56%) — trained with no pairs at all. The authors stress PRH is about positive-view quality, not scale, and invite readers to test the demo on their own photos/videos. Joint work with Yann LeCun, Brown, NYU, and amilabs.
More from Research
- After 4 years growing brain cells on electrodes, startup ships 0.5¢/sec video model mischief 1 — emmanuelvivier · 2026-10-31
- Steering an "Einstein neuron" in Qwen 4B produces dramatically different answers — thursdai_pod · 2026-10-31
- Protein binder scoring head-to-head: ipDAE consistently strong, OpenDDE+ipTM surprisingly good at 80% PPV — AllThingsApx · 2026-10-31
- Turing TTS benchmark: best voice AI fools listeners only ~10% of the time — omarsar0 · 2026-10-31
- Ablation study: fine-grained MoE is the best LLM architecture change since GPT-3 — stochasticchasm · 2026-10-31
- Yann LeCun to Keynote MAIN2026 on World Models for Reasoning and Planning — karimjerbineuro · 2026-10-31