Podcast

Best AI papers explained

Author: Enoch H. Kang
Feed episode count: 789
Official site: https://podcasters.spotify.com/pod/show/ehwkang
RSS: https://anchor.fm/s/1026675f8/podcast/rss
JSON: /v1/public/podcasts/best-ai-papers-explained-7258006
Markdown: /podcast/best-ai-papers-explained-7258006.md

Summary

Cut through the noise. We curate and break down the most important AI papers so you don’t have to.

Episodes

Jul 18, 2026 · not_requested

From Reasoning Traces to Reusable Modules: Understanding Compositional Generalization in Language Model Reasoning

Duration seconds: 1118
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/From-Reasoning-Traces-to-Reusable-Modules-Understanding-Compositional-Generalization-in-Language-Model-Reasoning-e3m89j9

HTML Markdown JSON

Jul 17, 2026 · not_requested

Position: Interpretability can be actionable

Duration seconds: 1482
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Position-Interpretability-can-be-actionable-e3m7963

HTML Markdown JSON

Jul 17, 2026 · not_requested

High-accuracy sampling for diffusion models and log-concave distributions

Duration seconds: 1349
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/High-accuracy-sampling-for-diffusion-models-and-log-concave-distributions-e3m6fc4

HTML Markdown JSON

Jul 15, 2026 · not_requested

Causal Inference with Video Features as Treatments

Duration seconds: 1333
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Causal-Inference-with-Video-Features-as-Treatments-e3m4i2f

HTML Markdown JSON

Jul 15, 2026 · not_requested

What Does Thompson Sampling Optimize?

Duration seconds: 1338
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/What-Does-Thompson-Sampling-Optimize-e3m3i4n

HTML Markdown JSON

Jul 13, 2026 · not_requested

Globally Convergent Offline Reinforcement Learning with Smoothed Bellman Residual Minimization

Duration seconds: 745
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Globally-Convergent-Offline-Reinforcement-Learning-with-Smoothed-Bellman-Residual-Minimization-e3m1h8c

HTML Markdown JSON

Jul 10, 2026 · not_requested

LLM-as-a-Verifier: A General-Purpose Verification Framework

Duration seconds: 1202
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/LLM-as-a-Verifier-A-General-Purpose-Verification-Framework-e3lt3rr

HTML Markdown JSON

Jul 9, 2026 · not_requested

How Much Do Language Models Memorize?

Duration seconds: 1432
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/How-Much-Do-Language-Models-Memorize-e3lrl61

HTML Markdown JSON

Jul 7, 2026 · not_requested

Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering

Duration seconds: 1307
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Position-Uncertainty-Quantification-in-LLMs-is-Just-Unsupervised-Clustering-e3lpdao

HTML Markdown JSON

Jul 6, 2026 · not_requested

Position: Agents Should Invoke External Tools ONLY When Epistemically Necessary

Duration seconds: 728
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Position-Agents-Should-Invoke-External-Tools-ONLY-When-Epistemically-Necessary-e3lo7do

HTML Markdown JSON

Jul 5, 2026 · not_requested

From conversations to mechanisms: aligning advertiser Incentives in ai-powered product recommendations

Duration seconds: 1322
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/From-conversations-to-mechanisms-aligning-advertiser-Incentives-in-ai-powered-product-recommendations-e3lmips

HTML Markdown JSON

Jul 4, 2026 · not_requested

Is one layer enough? Training a single transformer layer can match full-parameter RL training

Duration seconds: 1385
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Is-one-layer-enough--Training-a-single-transformer-layer-can-match-full-parameter-RL-training-e3llfe8

HTML Markdown JSON

Jul 2, 2026 · not_requested

RL Excursions during Pre-Training: Re-examining Policy Optimization for LLM training

Duration seconds: 1307
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/RL-Excursions-during-Pre-Training-Re-examining-Policy-Optimization-for-LLM-training-e3lja5t

HTML Markdown JSON

Jul 1, 2026 · not_requested

Language Generation with Feedback: Queries and Mistakes

Duration seconds: 1207
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Language-Generation-with-Feedback-Queries-and-Mistakes-e3lhnal

HTML Markdown JSON

Jul 1, 2026 · not_requested

Quantifying Theoretical AI Alignment Guarantees: Receiver-Utility Bounds in Bayesian Persuasion

Duration seconds: 1338
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Quantifying-Theoretical-AI-Alignment-Guarantees-Receiver-Utility-Bounds-in-Bayesian-Persuasion-e3lgi9d

HTML Markdown JSON

Jun 29, 2026 · not_requested

SPIRAL: Learning to search and aggregate

Duration seconds: 1335
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/SPIRAL-Learning-to-search-and-aggregate-e3lek6j

HTML Markdown JSON

Jun 27, 2026 · not_requested

Qwen-AgentWorld: Language World Models for General Agents

Duration seconds: 1244
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Qwen-AgentWorld-Language-World-Models-for-General-Agents-e3lbu3b

HTML Markdown JSON

Jun 27, 2026 · not_requested

When Does Trajectory-Level Supervision Permit Efficient Offline Reinforcement Learning?

Duration seconds: 1136
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/When-Does-Trajectory-Level-Supervision-Permit-Efficient-Offline-Reinforcement-Learning-e3lb91k

HTML Markdown JSON

Jun 26, 2026 · not_requested

SuperThoughts: Reasoning Tokens in Superposition

Duration seconds: 1140
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/SuperThoughts-Reasoning-Tokens-in-Superposition-e3lasbo

HTML Markdown JSON

Jun 25, 2026 · not_requested

First-Explore PPO : Learning Meta-Exploration with Proximal Policy Optimization

Duration seconds: 1355
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/First-Explore-PPO--Learning-Meta-Exploration-with-Proximal-Policy-Optimization-e3l998g

HTML Markdown JSON

Jun 24, 2026 · not_requested

Self-Distillation for Data-Scarce Language Model Pretraining

Duration seconds: 1305
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Self-Distillation-for-Data-Scarce-Language-Model-Pretraining-e3l70q8

HTML Markdown JSON

Jun 21, 2026 · not_requested

Meta-Harness for Agent-State Construction

Duration seconds: 1382
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Meta-Harness-for-Agent-State-Construction-e3l33v2

HTML Markdown JSON

Jun 21, 2026 · not_requested

ExpRL: Using Reference Solutions as Rewards for LLM Mid-Training

Duration seconds: 1263
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/ExpRL-Using-Reference-Solutions-as-Rewards-for-LLM-Mid-Training-e3l2mpb

HTML Markdown JSON

Jun 18, 2026 · not_requested

Valid Inference with Synthetic Data via Task Exchangeability

Duration seconds: 788
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/Valid-Inference-with-Synthetic-Data-via-Task-Exchangeability-e3l00o8

HTML Markdown JSON

Jun 17, 2026 · not_requested

GRPO is Secretly a Process Reward Model

Duration seconds: 1233
Source: https://podcasters.spotify.com/pod/show/ehwkang/episodes/GRPO-is-Secretly-a-Process-Reward-Model-e3kuhr6

HTML Markdown JSON