Podcast · 475 episodes
Best AI papers explained, page 2
by Enoch H. Kang · English
Cut through the noise. We curate and break down the most important AI papers so you don’t have to.
All episodes, page 2
Q-Learning with World ModelsQWM (Q-Learning with World Models) for faster, safer robotic learning by combining model-free Q-learning with short-horizon world-model lookahead to avoid compounding simulation bias.23 Aug 2026 · 25 min · 15 chapters
Conformal Language Modeling via Posterior SamplingConformal language modeling via posterior sampling, a MIT statistical method to reduce LLM hallucinations during text generation by conditioning token sampling on a calibrated “high-trust” region,…20 Aug 2026 · 22 min · 10 chapters
BoNVoyage: Learning Better Rewards without RankingExplains why RLHF reward models trained with pairwise rankings (Bradley-Terry) fail under adversarial distribution shift, enabling reward hacking, and presents Bon Voyage, a method to train reward…20 Aug 2026 · 22 min · 10 chapters
Demystifying Agent Skills: Why They Work—Until They Don’tWhy AI “agent skills” (distilled procedural checklists) improve performance, and when they fail.18 Aug 2026 · 20 min · 12 chapters
Jagged Judges: Epistemic Stability Under Silence, Pressure, and PersistenceEpistemic stability of AI “judges” (LLMs used for grading, moderation, and reward modeling) under silence, pressure, and persistence; argues that standard accuracy tests are misleading because models…15 Aug 2026 · 20 min · 11 chapters
Predicting Neural Scaling Laws without Training: A Data Manifold OracleThe episode explains the “Data Manifold Oracle” (DMO), a training-free method to predict neural scaling laws (the “floor” error limit and the “slope” learning rate) from raw text using standard…15 Aug 2026 · 22 min · 12 chapters
Making Interpretable Discoveries from Unstructured Data: A High-Dimensional Multiple Hypothesis TestingA July 2026 research framework for making interpretable, bias-resistant discoveries from unstructured text using sparse autoencoders, high-dimensional multiple-hypothesis testing (KFWER), Gaussian…11 Aug 2026 · 24 min · 14 chapters
Overcoming the Incentive Collapse ParadoxThe “incentive collapse paradox” in AI-human workflows: if pay depends only on final accuracy, near-perfect AI makes workers rationally free-ride (exert less effort), potentially requiring infinite…11 Aug 2026 · 21 min · 11 chapters
Position: Modular Memory is the Key to Continual Learning AgentsContinual learning agents that learn over a lifetime without catastrophic forgetting, using modular memory (core model + working memory + long-term memory) and two regimes: external interaction and…10 Aug 2026 · 27 min · 14 chapters
Harness RL is Meta-Learning: Training to Self-Improve at Test TimeHarness RL for meta-learning/self-improvement at test time by moving adaptation from updating model weights to revising the “harness” (instructions, memory, tool rules, verification).8 Aug 2026 · 22 min · 9 chapters
Escaping the Nash Trap: Structural Estimation and Alignment of Strategic Reasoning in Large Language ModelsThe episode argues that large language models often fail in strategic settings because they assume opponents are perfectly rational “Nash-type” optimizers, creating “Nash traps” where the…7 Aug 2026 · 22 min · 10 chapters
When Does LeJEPA Learn a World Model?When LeJEPA (Joint Embedding Predictive Architecture) learns a reliable “world model,” i.e., linearly identifiable latent variables from pixel observations, and when that guarantee fails in real…7 Aug 2026 · 23 min · 17 chapters
Do Modules Stay in Their Lane? Role Drift in Compound LLM SystemsThe episode explains “role drift” in compound LLM systems trained with outcome-only reinforcement learning, where modules abandon their intended roles while still achieving high final accuracy.3 Aug 2026 · 21 min · 13 chapters
Do you really need to pretrain Q-functions for online RL fine-tuning?Whether online reinforcement learning fine-tuning needs a pre-trained Q-function (critic) and why naive offline Q pretraining can hurt.1 Aug 2026 · 24 min · 10 chapters
The Evolution of Digital Search: From Blue Links to Delegated Decision-MakingThe episode argues that traditional keyword “blue link” search is dying and being replaced by AI-native delegated decision-making, where consumer agents interpret intent, evaluate options invisibly,…29 Jul 2026 · 19 min · 8 chapters
Ask, Don’t Judge: Binary Questions for Interpretable LLM Evaluation and Self-ImprovementHow to evaluate LLM outputs reliably without opaque “holistic” scores, using binary yes/no checklists (Binavol) to catch specific errors and enable automated self-correction.28 Jul 2026 · 5 min · 4 chapters
Understanding Reasoning from Pretraining to Post-TrainingHow reasoning emerges from the standard AI pipeline (pretraining + supervised fine-tuning + reinforcement learning), using chess and math as controlled testbeds; includes scaling laws, compute…24 Jul 2026 · 22 min · 12 chapters
A Positive Case for Faithfulness: LLM Self-Explanations Help Predict Model BehaviorWhether LLM “self-explanations” (the reasons they give alongside answers) are faithful to their actual internal decision logic, and whether explanations improve prediction of future behavior.23 Jul 2026 · 15 min · 9 chapters
Reject, Resample, Repeat: Understanding Parallel Reasoning in Language Model InferenceHow inference-time “parallel reasoning” works for language models using sequential Monte Carlo (SMC) with process reward models (PRMs), including math guarantees and why theory can fail on real…19 Jul 2026 · 23 min · 11 chapters
Rethinking the Evaluation of Harness Evolution for AgentsEvaluates “automatic harness evolution” for AI agents—whether letting a model rewrite its prompts/tools/control logic actually makes it smarter, or just exploits benchmark evaluation.19 Jul 2026 · 23 min · 13 chapters
From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model ReasoningCompositional generalization in language-model reasoning—why supervised fine-tuning (SFT) can make models rigid, and how reinforcement learning (RL) “untangles” reasoning into reusable skill and…18 Jul 2026 · 19 min · 11 chapters
Position: Interpretability can be actionableThe episode argues that AI interpretability must become actionable—moving from passive “black box” observation to concrete interventions that improve real-world behavior, safety, and usability.17 Jul 2026 · 25 min · 12 chapters
High-accuracy sampling for diffusion models and log-concave distributionsA 2026 theoretical breakthrough on high-accuracy sampling for diffusion models and log-concave distributions using only first-order information (gradients), avoiding the “polynomial wall” of standard…17 Jul 2026 · 22 min · 9 chapters
Causal Inference with Video Features as TreatmentsCausal inference for video persuasion—isolating the frame-by-frame causal effect of a fleeting visual feature (treated as a “treatment”) on changing human emotion, while correcting for confounding…15 Jul 2026 · 22 min · 11 chapters