Center for Language & Speech Processing(CLSP), JHU Courses

Controllable Safety Alignment - Inference-Time Adaptation to Diverse Safety Requirements

Explore how language models can adapt to diverse safety requirements through Controllable Safety Alignment, enabling flexible content moderation across different cultural contexts and user needs.

• YouTube
• 52 minutes
• Self-Paced
• Free Video

Using Semantics to Help Learn Phonetic Categories

Discover how semantics can enhance phonetic category learning through computational models and linguistic analysis in this research-focused exploration of language acquisition.

• YouTube
• 51 minutes
• Self-Paced
• Free Video

AppWorld: Reliable Evaluation of Interactive Agents in a World of Apps and People

Explore AppWorld's innovative framework for evaluating AI agents' performance in everyday digital tasks across popular apps, highlighting challenges and future directions for reliable AI systems.

• YouTube
• 1 hour 9 minutes
• Self-Paced
• Free Video

Probabilistic Inference in Language Models via Twisted Sequential Monte Carlo

Explore advanced probabilistic inference techniques for Large Language Models using Sequential Monte Carlo methods, focusing on safety, capability enhancement, and evaluation frameworks.

• YouTube
• 1 hour 1 minute
• Self-Paced
• Free Video

Improving LLM Generalization by Selecting and Synthesizing Data

Discover advanced techniques for enhancing language model performance through strategic data selection and synthesis, focusing on cross-task generalization and domain adaptation in NLP.

• YouTube
• 59 minutes
• Self-Paced
• Free Video

Target Sound Extraction with Language-oriented Audio Diffusion Transformer

Discover how SoloAudio uses diffusion models and transformers to extract specific sounds from audio, with innovative language-oriented capabilities and zero-shot learning features.

• YouTube
• 31 minutes
• Self-Paced
• Free Video

Self-Incorrect - LLMs Struggle with Discriminating Self-Generated Responses

Explore groundbreaking research on LLMs' limitations in self-improvement, revealing their struggles to effectively discriminate and enhance their own generated responses.

• YouTube
• 12 minutes
• Self-Paced
• Free Video

Using Language Models for Instruction-Based Information Retrieval

Explore instruction-based information retrieval using language models. Learn about Promptriever, a new retriever showing significant gains in instruction following and promptability.

• YouTube
• 40 minutes
• Self-Paced
• Free Video

Takeaways from the SCALE 2024 Workshop on Video-based Event Retrieval

Explore multilingual event-centric video retrieval techniques, focusing on non-professional content. Learn about dataset creation, modality-specific improvements, and key findings in text extraction and fusion.

• YouTube
• 1 hour 7 minutes
• Self-Paced
• Free Video

End-to-End Neural Speaker Diarization with Non-Autoregressive Attractors

Explore end-to-end neural speaker diarization with non-autoregressive attractors, improving robustness for real-life scenarios and handling variable speaker numbers in recordings.

• YouTube
• 37 minutes
• Self-Paced
• Free Video

Efficient Speech Processing

Explore cutting-edge research in speech recognition, generation, and privacy. Learn about self-training algorithms, error correction models, joint text-to-speech modeling, and privacy-preserving techniques in speech processing.

• YouTube
• 1 hour 8 minutes
• Self-Paced
• Free Video

Extension is All You Need - But Sometimes Good Intentions Help Too

Explore innovative approaches to language extension and their impact on linguistic analysis, with insights on balancing intentions and outcomes.

• YouTube
• 1 hour
• Self-Paced
• Free Video

OLMo: Accelerating the Science of Open Language Models

Explore OLMo, an open-source language model initiative, and its impact on accelerating research in natural language processing and machine learning.

• YouTube
• 1 hour 3 minutes
• Self-Paced
• Free Video

A Feature-Based Model of Human Lexical Access - 2004

Explore a feature-based model of human lexical access, delving into the intricacies of language processing and speech recognition.

• YouTube
• 1 hour 13 minutes
• Self-Paced
• Free Video

Towards Meaning: The Tectogrammatical Sentence Representation - 2001

Explore tectogrammatical sentence representation, a linguistic approach for analyzing sentence structure and meaning, presented by Jan Hajic in 2001.

• YouTube
• 1 hour 19 minutes
• Self-Paced
• Free Video