Taewoon KimAI researcher and engineer.https://taewoon.kim/My Neuro-Symbolic Meta-Policies paper was accepted at ISWC 2026https://taewoon.kim/2026/07/28/roomkg-meta-policies-iswc-2026/https://taewoon.kim/2026/07/28/roomkg-meta-policies-iswc-2026/Why I want agents to choose their memory rules, not hide themTue, 28 Jul 2026 00:00:00 GMTneuro-symbolic AIknowledge graphreinforcement learningTaewoon KimMy Co-Learning paper was accepted at RO-MAN 2026https://taewoon.kim/2026/06/19/co-learning-ro-man-2026/https://taewoon.kim/2026/06/19/co-learning-ro-man-2026/Why I want a robot to remember how a team worked beforeFri, 19 Jun 2026 00:00:00 GMThuman-robot interactionknowledge graphepisodic memoryTaewoon KimMy KG Memory Transfer paper was accepted at RLC 2026https://taewoon.kim/2026/05/22/kg-memory-transfer-rlc-2026/https://taewoon.kim/2026/05/22/kg-memory-transfer-rlc-2026/Why I care about learning which symbolic facts should persistFri, 22 May 2026 00:00:00 GMTreinforcement learningknowledge graphneuro-symbolic AITaewoon KimBuilding embedded Python bindings for ArcadeDBhttps://taewoon.kim/2026/01/28/building-arcadedb-embedded-python-bindings/https://taewoon.kim/2026/01/28/building-arcadedb-embedded-python-bindings/Why I wanted a lightweight, local-first way to use ArcadeDB from PythonWed, 28 Jan 2026 00:00:00 GMTArcadeDBPythonopen sourcedatabasesgraphsembeddingsJVMhumemaiTaewoon KimWhat Is a Knowledge Graph? A Practical Guide Across RDF and Property Graphshttps://taewoon.kim/2025/10/06/knowledge-graph/https://taewoon.kim/2025/10/06/knowledge-graph/Formal basics, real systems, and why industry favors property graphs while RDF remains importantMon, 06 Oct 2025 00:00:00 GMTknowledge-graphsrdfproperty-graphssemantic-webneo4jjenaneptunegraphdbdata-engineeringTaewoon KimFrom REINFORCE to PPO: The Complete On-Policy RL Journeyhttps://taewoon.kim/2025/08/07/on-policy-rl/https://taewoon.kim/2025/08/07/on-policy-rl/Understanding the evolution from basic policy gradients to modern LLM fine-tuning algorithmsThu, 07 Aug 2025 00:00:00 GMTreinforcement-learningpolicy-gradientsppogrpollm-trainingdeep-learningpytorchtutorialTaewoon KimFrom VAEs to Diffusion Models: A Step-by-Step Journeyhttps://taewoon.kim/2025/05/28/vae-diffusion/https://taewoon.kim/2025/05/28/vae-diffusion/Understanding the evolution of generative models through practical implementationsWed, 28 May 2025 00:00:00 GMTvaediffusion-modelsgenerative-aideep-learningpytorchtutorialTaewoon KimDiscrete vs. Continuous: A Tale of Two Approaches to Language Modelinghttps://taewoon.kim/2025/03/09/discrete-vs-continuous/https://taewoon.kim/2025/03/09/discrete-vs-continuous/Balancing n-gram models and neural networks for next-token predictionSun, 09 Mar 2025 00:00:00 GMTlanguage-modelingdiscrete-searchneural-networkschain-of-thoughtn-gramTaewoon KimRL vs SL: Understanding Their Roles in Large Language Modelshttps://taewoon.kim/2025/02/06/sl-vs-rl-gradients/https://taewoon.kim/2025/02/06/sl-vs-rl-gradients/Bridging Reinforcement Learning and Supervised Learning in LLMsThu, 06 Feb 2025 00:00:00 GMTreinforcement-learningsupervised-learningllmmachine-learningRLSLTaewoon KimUnderstanding Maximum Likelihood Estimationhttps://taewoon.kim/2025/02/05/mle/https://taewoon.kim/2025/02/05/mle/A Deep Dive into MLE, Loss Functions, and BeyondWed, 05 Feb 2025 00:00:00 GMTstatisticsmachine-learningestimationlikelihoodTaewoon KimSequential Decision-Making with Transformershttps://taewoon.kim/2024/12/15/sequential/https://taewoon.kim/2024/12/15/sequential/Offline RL, Behavior Cloning, and The Magic of Sequence ModelingSun, 15 Dec 2024 00:00:00 GMTtransformersoffline-rlsequence-modelingdecision-transformerTaewoon KimFrom 1+1 in Assembly to LLMs: The Evolution of Computing Abstractionhttps://taewoon.kim/2024/11/12/1+1/https://taewoon.kim/2024/11/12/1+1/Tracing the Layers from Machine Code to Natural Language InterfacesTue, 12 Nov 2024 00:00:00 GMTcomputingprogramming languagesabstraction layersassemblyCJavaPythonLLMsTaewoon KimThe Problems with p-valueshttps://taewoon.kim/2024/10/16/p-value-bayesian/https://taewoon.kim/2024/10/16/p-value-bayesian/Why Frequentist Significance Testing Falls ShortWed, 16 Oct 2024 00:00:00 GMTstatisticsp-valuesdata sciencefrequentist vs bayesianstatistical significancehypothesis testingTaewoon KimCan the Transformer be viewed as a special case of a Graph Neural Network (GNN)?https://taewoon.kim/2024/10/15/transformer-vs-gnn/https://taewoon.kim/2024/10/15/transformer-vs-gnn/A natural language text can be seen as a knowledge graphTue, 15 Oct 2024 00:00:00 GMTGNNtransformerknowledge graphNLPnatural languageTaewoon KimIs supervised learning a special type of reinforcement learning?https://taewoon.kim/2024/09/22/rl-sl/https://taewoon.kim/2024/09/22/rl-sl/The reason why reinforcement learning is such a hard problemSun, 22 Sep 2024 00:00:00 GMTreinforcement learningsupervised learningcontextual banditsTaewoon KimPlaying around with Hugging Face Llama 3.1 locallyhttps://taewoon.kim/2024/08/29/llama/https://taewoon.kim/2024/08/29/llama/4bit quantization is amazingThu, 29 Aug 2024 00:00:00 GMTllmllamahugging faceTaewoon KimTraining a GCN-based edge classifierhttps://taewoon.kim/2024/08/21/gnn-edge-classifier/https://taewoon.kim/2024/08/21/gnn-edge-classifier/Exploring the challenges and strategies for effective edge classification with graph neural networks (GNNs)Wed, 21 Aug 2024 00:00:00 GMTknowledge graphmachine learningdeep learninggnnTaewoon KimA machine with human-like memory systemshttps://taewoon.kim/2023/12/21/human-memory/https://taewoon.kim/2023/12/21/human-memory/Can machines think like us?Thu, 21 Dec 2023 00:00:00 GMTreinforcement learningAIhuman memoryknowledge graphmachine learningdeep learningPhDLLMsymbolic AIGOFAIhumemaiTaewoon KimUsing ChatGPT is just so greathttps://taewoon.kim/2023/12/14/chatgpt/https://taewoon.kim/2023/12/14/chatgpt/This really can do a lot of things, although it's still biased.Thu, 14 Dec 2023 00:00:00 GMTChatGPTAIGPT4DALL-ETaewoon KimMy new websitehttps://taewoon.kim/2023/12/12/new-website/https://taewoon.kim/2023/12/12/new-website/My boring old website had to be gone.Tue, 12 Dec 2023 00:00:00 GMTwebsiteTaewoon KimBloom: A 176b-parameter open-access multilingual language modelhttps://taewoon.kim/2022/11/09/bloom/https://taewoon.kim/2022/11/09/bloom/Unlocking the power of open-source language modelsWed, 09 Nov 2022 00:00:00 GMTllmpromptgptbloomTaewoon KimRoom classifierhttps://taewoon.kim/2022/02/28/room/https://taewoon.kim/2022/02/28/room/A simple room classifier made using EfficientNet and PyTorch LightningMon, 28 Feb 2022 00:00:00 GMTroomclassifierTaewoon KimPromptsource: An integrated development environment and repository for natural language promptshttps://taewoon.kim/2022/02/02/promptsource/https://taewoon.kim/2022/02/02/promptsource/Revolutionizing NLP with a collaborative hub for crafting and sharing promptsWed, 02 Feb 2022 00:00:00 GMTllmpromptgptpromptsourceTaewoon KimIGLU: interactive grounded language understanding in a collaborative environmenthttps://taewoon.kim/2021/12/16/iglu/https://taewoon.kim/2021/12/16/iglu/We won the NeurIPS 2021 competitionThu, 16 Dec 2021 00:00:00 GMTIGLUreinforcement learningNeurIPScompetitionTaewoon KimMultitask prompted training enables zero-shot task generalizationhttps://taewoon.kim/2021/10/15/multi-task/https://taewoon.kim/2021/10/15/multi-task/Achieving superior zero-shot generalization through explicit multitask promptingFri, 15 Oct 2021 00:00:00 GMTllmpromptgptTaewoon KimEmoBERTa: Speaker-aware emotion recognition in conversation with RoBERTahttps://taewoon.kim/2021/08/26/emoberta/https://taewoon.kim/2021/08/26/emoberta/Achieving state-of-the-rrt emotion recognition in conversations with a simple RoBERTa-based approachThu, 26 Aug 2021 00:00:00 GMTllmbertrobertaemotionercTaewoon KimGeneralizing MLPs with dropouts, batch normalization, and skip connectionshttps://taewoon.kim/2021/08/18/age-gender/https://taewoon.kim/2021/08/18/age-gender/A simple age-gender classification model using Arcface embeddings and MLPsWed, 18 Aug 2021 00:00:00 GMTmlpagegenderTaewoon Kim