TechcraftingAI NLP

TechcraftingAI NLP

By Brad EdwardsTechnology
Download on the App Store

All TechcraftingAI NLP episodes

  • Ep. 41 - November 3, 2023

    arXiv research summaries for Computation and Language from November 3, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


    Today's research themes (AI summary)

    • Developing methods to make large language models more interpretable and explainable
    • Studying prompt engineering techniques to improve few-shot learning with large language models
    • Evaluating cross-lingual transfer learning methods for low-resource languages
    • Analyzing model biases and proposing techniques to mitigate demographic and framing biases
    • Benchmarking large language models on complex multimodal tasks like editing PowerPoint slides
    • 42 min
    • Ep. 40 - November 2, 2023

      arXiv research summaries for Computation and Language from November 2, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


      Today's research themes (AI summary)

      • Enhancing NLP models with external knowledge through techniques like prompting and knowledge injection.
      • Analyzing model performance on challenging linguistic examples that reveal nuances like syntactic dependencies or semantic inconsistencies.
      • Developing methods to control model behavior, like refusal mechanisms and evaluating consistency.
      • Leveraging generative capabilities for applications like data augmentation and topic modeling.
      • Adapting models to new domains/languages, via transfer learning and multitask training.
      • 34 min
      • Ep. 39 - November 1, 2023

        arXiv research summaries for Computation and Language from November 1, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠.


        Today's research themes (AI summary)

        • Evaluating and improving biases and safety issues in large language models.
        • Using large language models for summarization and generation.
        • Developing better evaluation metrics for language tasks.
        • Applying large language models to multilingual and low-resource languages.
        • Improving language models with explicit linguistic knowledge.
        • 37 min
        • Ep. 38 - October 31, 2023

          arXiv research summaries for Computation and Language from October 31, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠.


          Today's research themes (AI summary)

          • Improving reasoning abilities of large language models through techniques like learning from mistakes.
          • Evaluating capabilities of large language models through games, analogies, and psychological assessments.
          • Building better sentiment analysis systems using non-compositional phrases.
          • Transporting causal effects between text domains for valid causal inference.
          • Analyzing contents of large text corpora for quality, inclusion, and contamination.
          • 1 hr 5 min
          • Ep. 37 - October 30, 2023

            arXiv research summaries for Computation and Language from October 30, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠.


            Today's research themes (AI summary)

            • Improving natural language processing through novel large language model architectures and training techniques.
            • Using large language models for knowledge extraction and retrieval applications.
            • Evaluating and analyzing the capabilities and limitations of large language models.
            • Leveraging large language models for personalized systems and recommendations.
            • Applying large language models to domain-specific language tasks like clinical NLP and biomedical NLP.
            • 1 hr 5 min
            • Ep. 36 - October 29, 2023

              arXiv research summaries for Computation and Language from October 29, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠.


              Today's research themes (AI summary)

              • Mitigating bias in large language models.
              • Improving emotion and etiquette understanding in LLMs.
              • Novel model architectures for complex named entity recognition.
              • Ensembling and stacking models for improved AI text detection.
              • Simplifying the smart reply pipeline with end-to-end retrieval.
              • 8 min
              • Ep. 35 - October 28, 2023

                arXiv research summaries for Computation and Language from October 28, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠.


                Today's research themes (AI summary)

                • Improving efficiency and reducing computational costs of large language models through methods like early exiting and knowledge distillation.
                • Developing new datasets and models for low-resource languages and cross-lingual transfer, including for tasks like speech recognition and machine translation.
                • Detecting and mitigating biases in language models, including human-like biases that propagate from pre-trained models to downstream tasks like speech emotion recognition.
                • Advancing narrative and relational understanding capabilities of language models through new tasks, datasets and modular approaches.
                • Enhancing reasoning across modalities in language models, with a focus on integrating visual information from images with text for knowledge extraction and medical question answering.
                • 32 min
                • Ep. 34 - October 27, 2023

                  arXiv research summaries for Computation and Language from October 27, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠.


                  Today's research themes (AI summary)

                  • Large Language Models: Pushing the Limits of Understanding
                  • Code-Switching NLP: New Frontiers in Multilingual AI
                  • Exposing Misinformation: Automated Fact-Checking Techniques
                  • Conversational Agents Get a Makeover: Personas and Persuasion
                  • Reading Between the Lines: Advances in Machine Comprehension
                  • 44 min
                  • Ep. 33 - October 26, 2023

                    arXiv research summaries for Computation and Language from October 26, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠.


                    Today's research themes (AI summary)

                    • Advances in large language models through fine-tuning, prompting, and quantization. Examples include HuggingFace LLaMA, custom prompting for ChatGPT, and quantizing models down to 4 bits.
                    • Analyzing capabilities and limitations of large language models on reasoning, commonsense, and factual knowledge tasks. Papers probe model knowledge in areas like theory of mind, spatial reasoning, and stumpers.
                    • Improving natural language processing applications with techniques like contrastive learning, targeted pretraining, and reinforcement learning from human feedback. Areas include intent detection, question answering, summarization.
                    • Evaluating bias, fairness, transparency, and ethics of large language models and datasets. Analyzing issues like occupational stereotypes, licensing and attribution of training data, and fairness-explainability tradeoffs.
                    • Enabling multilinguality, cross-linguality, and transfer learning. Examples include cross-lingual transfer for low-resource languages, code embeddings across programming languages, and continual learning for multilingual ASR.
                    • 52 min
                    • Ep. 32 - October 25, 2023

                      arXiv research summaries for Computation and Language from October 25, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠.


                      Today's Theme (LLM-Generated)

                      • Evaluating capabilities and limitations of large language models (LLMs) like ChatGPT and GPT-4 through tasks such as reasoning, summarization, and parsing.
                      • Improving alignment and control of LLMs through techniques like reinforcement learning from human feedback, supervised iterative learning, and contextualized prompt optimization.
                      • Enhancing fairness, interpretability, and transparency of models through bias mitigation, rationale extraction, and pretraining data detection.
                      • Advancing multilingual and multimodal models through pretraining, benchmarking, and adapting models like mBERT to new languages and modalities.
                      • Quantization and efficiency improvements for deploying large models through methods like post-training 4-bit floating point quantization.
                      • 1 hr 4 min