TechcraftingAI NLP

TechcraftingAI NLP

By Brad EdwardsTechnology
Download on the App Store

All TechcraftingAI NLP episodes

  • Ep. 79 - December 12, 2023

    arXiv NLP research summaries for December 13, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


    Today's research themes (AI summary)

    • Improving factual accuracy of large language models through alignment techniques or error injection
    • Mitigating aggressive fine-tuning of pretrained models with limited labeled data
    • Generating coherent narratives for visual scenes using diffusion models
    • Adapting large language models to scientific domains with contextual calibration
    • Extracting relational knowledge from text through bi-encoder-decoder models
    • 32 min
    • Ep. 78 - December 11, 2023

      arXiv NLP research summaries for December 11, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


      Today's research themes (AI summary)

      • Mitigating toxicity and bias in large language models while preserving performance
      • Enhancing few-shot learning capabilities for low-resource languages and tasks
      • Developing explainable and transparent approaches for evaluating model behavior
      • Advancing multimodal learning through combined use of text, images, audio and networks
      • Creating retrievers and decoders for more focused, relevant, and fact-based generation
      • 25 min
      • Ep. 77 - December 10, 2023

        arXiv NLP research summaries for December 10, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


        Today's research themes (AI summary)

        • Compressing and distilling large language models for efficiency
        • Applying large language models to open-domain fact checking
        • Incorporating commonsense knowledge into language models
        • Using language models for lexical semantic change detection
        • Developing multimodal models for meme analysis
        • 16 min
        • Ep. 76 - December 9, 2023

          arXiv NLP research summaries for December 9, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


          Today's research themes (AI summary)

          • Analyzing and detecting toxic language in social media
          • Understanding and improving team collaboration through dialogue analysis
          • Using transformer models like BERT for text classification tasks
          • Explaining predictions of generative language models
          • Editing knowledge in large language models while retaining historical knowledge
          • 20 min
          • Ep. 75 - December 8, 2023

            arXiv NLP research summaries for December 8, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


            Today's research themes (AI summary)

            • Protecting copyright of large language models with human-readable fingerprints
            • Creating datasets and models for dialogue summarization in French
            • Using retrieval to augment reasoning in multi-agent debates
            • Generating explanations to understand and repair entity alignment models
            • Leveraging text retrieval to enhance predictive chemistry models
            • 28 min
            • Ep. 74 - December 7, 2023

              arXiv NLP research summaries for December 7, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


              Today's research themes (AI summary)

              • Developing multimodal models for detecting misinformation in social media
              • Evaluating language models for providing personalized coaching messages for behavioral weight loss
              • Analyzing the tendency of language models to break safety mechanisms when given malicious instructions
              • Applying knowledge distillation to create efficient question answering models for Spanish
              • Probing language models across scales and layers to understand their reasoning abilities
              • 27 min
              • Ep. 73 - December 6, 2023

                arXiv NLP research summaries for December 6, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


                Today's research themes (AI summary)

                • Improving cross-lingual speech recognition through better phoneme recognition and grapheme translation
                • Analyzing limitations of generic toxicity classifiers for detecting misogyny
                • Exploring the role of topic and genre in dialogue systems
                • Techniques to teach scientific knowledge to large language models
                • Scaling natural language interfaces for querying massive databases
                • 25 min
                • Ep. 72 - December 5, 2023

                  arXiv NLP research summaries for December 5, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


                  Today's research themes (AI summary)

                  • Empathy, emotion, and personality detection in text
                  • Multilingual model ensembles for intimacy prediction
                  • Tokenization impact on Russian language model adaptation
                  • Compositional generalization for data-to-text generation
                  • Applying large language models to graph data
                  • 14 min
                  • Ep. 71 - December 4, 2023

                    arXiv NLP research summaries for December 4, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


                    Today's research themes (AI summary)

                    • Evaluating language models on ability to generate educational math questions and solve programming problems.
                    • Techniques to improve language model reasoning like providing additional context, communication between models, and self-critiquing outputs.
                    • Methods to detect machine generated text and model grounding abilities.
                    • Understanding media coverage patterns and news ecosystem dynamics with article similarity models.
                    • Leveraging open source code to train performant and unbiased language models for code.
                    • 23 min
                    • Ep. 70 - December 3, 2023

                      arXiv NLP research summaries for December 3, 2023. You can find summaries and links to each article ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠here⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.


                      Today's research themes (AI summary)

                      • Improving named entity recognition for low-resource languages like Marathi.
                      • Using explicitations to bridge knowledge gaps and improve machine translation.
                      • Benchmarking generative language models for Norwegian language tasks.
                      • Generating educational Arabic crossword puzzles with AI.
                      • Evaluating the quality of sentence splitting and rephrasing.
                      • 19 min