Abstracts: NeurIPS 2024 with Weizhu Chen

Channel:
Microsoft Research
Subscribers:
351,000
Published on ● Video Link: https://www.youtube.com/watch?v=EIdZgX9Daic


Duration: 0:00
31 views
0

Next-token prediction trains a language model on all tokens in a sequence. VP Weizhu Chen discusses his team’s 2024 NeurIPS paper on how distinguishing between useful and “noisy” tokens in pretraining can improve token efficiency and model performance.

Show notes: https://www.microsoft.com/en-us/research/podcast/abstracts-neurips-2024-with-weizhu-chen/

Listen to the Abstracts series: https://www.microsoft.com/en-us/research/podcast-series/abstracts/



Other Videos By Microsoft Research


2025-06-03Episode 7: The road ahead
2025-06-03Episode 6: Healthcare Agent Orchestrator
2025-06-03Episode 5: Breakthroughs in AI
2025-06-03Episode 4: A distribution channel for AI innovation
2025-06-03Episode 3: Collaborating faster
2025-06-03Episode 2: A multi-disciplinary approach
2025-06-03Episode 1: Tackling complex healthcare challenges
2025-06-03AI Frontiers: Rethinking intelligence with Ashley Llorens and Ida Momennejad
2025-06-03Abstracts: March 21, 2024
2025-06-03Ideas: Language technologies for everyone with Kalika Bali
2025-06-02Abstracts: NeurIPS 2024 with Weizhu Chen
2025-05-28What AI's impact on individuals means for the health workforce and industry
2025-05-21Abstracts: Zero-shot models in single-cell biology with Alex Lu
2025-05-21Ideas: AI for materials discovery with Tian Xie and Ziheng Lu
2025-05-21Ideas: Bug hunting with Shan Lu
2025-05-21Ideas: Building AI for population-scale systems with Akshay Nambi
2025-05-21Ideas: Quantum computing redefined with Chetan Nayak
2025-05-21The reality of generative AI in the clinic
2025-05-21Ideas: Accelerating Foundation Models Research: AI for all
2025-05-21Real-world healthcare AI development and deployment—at scale
2025-05-21Empowering patients and healthcare consumers in the age of generative AI