PODCAST · technology
Best AI papers explained
by Enoch H. Kang
Cut through the noise. We curate and break down the most important AI papers so you don’t have to.
Episodes
-
Hybrid Reasoning Agents
This paper introduces HybridAgents, a novel framework that blends continuous latent reasoning with explicit text interactions within a unified generation framework. By utilizing a lightweight adapter called HybridLink, the system seamlessly transitions between internal thinking modes and external...
-
Strategic Self-Consistency
This research demonstrates that unfaithful large language model service providers can artificially inflate the number of reasoning paths generated during test-time compute to overcharge users. By utilizing a specific algorithm to generate and strategically reorder extra paths, a provider can make...
-
Retrofitting language models to operate over bytes
This Nature paper discusses Byteification, which offers a novel method to convert traditional subword-based large language models into efficient byte-level models that handle raw text encodings natively. By introducing a two-stage training procedure and a latent tokenizer architecture with...
-
Priced Guidance: Can Language Models Generate Future Research Ideas?
This paper introduces Priced Guidance, a novel evaluation framework that assesses language models on their ability to generate novel and future research ideas. Rather than using expensive repeated sampling or verbatim text compression, this methodology measures the compression cost—the exact…
more -
Context Language Models
This paper introduce Context Language Models (CLMs), a novel approach that enables language models to natively manage their own context by treating it as an editable file rather than relying on traditional append-only histories or external harnesses. By granting models unrestricted read and write...
-
Position: AI-Agent Pricing Should Become More Outcome-Dependent: An Economic Perspective
This paper analyzes AI-agent pricing models through an economic lens, arguing that dominant token-based pricing is insufficient because it fails to share risk efficiently or incentivize hidden quality efforts like model selection and verification. By utilizing stylized principal-agent models, the...
-
Emergent Capabilities Arise Randomly from Learning Sparse Attention Patterns
This paper investigates why transformer language models display sudden, emergent capabilities during training by analyzing how they learn attention patterns. Through experiments on both real language models and synthetic tasks like linear maps and cellular automata, the authors demonstrate that...
-
Pragmatic Causal Inference with AI-Learned Representations
This paper presents a rigorous framework for performing valid causal inference when complex covariates are compressed into AI-learned representations or embeddings. The authors demonstrate that information loss from compression distorts causal parameters through a product of regression and...
-
Pointwise or pairwise: when do pairwise losses help reward learning, provably?
This research paper investigates when and why pairwise losses outperform pointwise losses for reward learning by analyzing both approaches within a grouped offline contextual-bandit framework. The authors compare Value Regression (VR), which directly fits absolute observed rewards, with Value...
-
When to Trust the Cheap Check: Weak and Strong Verification for Reasoning
This paper explores weak-strong verification policies for large language models, presenting a framework that balances the affordability of scalable internal checks with the precision of resource-intensive external validation. To address the trade-off between type-I errors, type-II errors, and...
-
Theoretical Limits of Language Model Alignment
This paper investigate the fundamental limits of language model alignment by establishing exact theoretical boundaries for reward improvement under a KL-divergence constraint using Jeffreys divergence and a computable covariance estimator. They demonstrate that best-of-N sampling closely…
more -
DT2: Decision-Targeted Digital Twins
This paper introduces DT2, a novel training framework designed to align digital twins more effectively with their primary goal of decision support. Traditional virtual models often fail to rank policy options correctly because they prioritize minimizing overall simulation errors rather than...
-
Self-Play Pretraining with Zero Data
This paper introduces Self-Play Pretraining with Zero Data, a method for training language models using only synthetic data generated by the model itself. In this framework, a generator creates programs for a universal Turing machine while a learner is trained to predict the resulting byte...
-
Language models need sleep: learning to self-modify and consolidate memories
This research paper introduces a "Sleep" paradigm for Large Language Models (LLMs) to overcome the limitations of static knowledge and catastrophic forgetting. Inspired by human biology, the framework alternates between active phasesfor processing external data and sleep phases for internal...
-
Jev Creator: System One models for Prod, not God
We discuss the Latent Space's interview with Diogo Almeida, CEO of TypeSafe, about the launch of Jev, a specialized class of AI models designed for programmatic integration rather than human conversation. He argues that traditional models are "fractured" by safety alignments and chatbot...
-
Detecting and countering misuse of AI: September 2026
This paper is an Anthropic threat intelligence report from September 2026 detailing the misuse of AI models by various malicious actors. It describes how state-sponsored groups and cybercriminals have integrated Claude into autonomous attack frameworks to accelerate the development of exploits…
more -
Position: LLMs can’t jump
This paper examines the cognitive limitations of Large Language Models by using Albert Einstein’s discovery of General Relativity as a primary case study. While modern AI excels at induction through data compression and deduction via logical proof, the author argues that it lacks the capacity for...
-
Jailbreaking Jailbreaks: A Proactive Defense for LLMs
The research introduces PROACT, a proactive defense framework designed to safeguard Large Language Models from iterative adversarial attacks. Unlike traditional passive defenses that offer standard refusals, this system generates spurious responses that mimic successful jailbreaks while remaining...
-
When Agents Slow Down: Understanding LLM Agents’ Test-Time Strategies via Elo-per-token Analysis
This paper introduces Elo-per-token analysis, a novel framework for measuring how the performance of large language model agents scales with increased inference-time computation. By analyzing diverse benchmarks, the authors demonstrate that while agents initially show efficiency gains, their...
-
Thinking with Looped Flows
This paper introduces looped flows, a novel framework designed to enhance the reasoning capabilities of neural networks by merging recurrent hidden states with probability flow models. Traditional looped models often struggle with training instability because they cannot effectively backpropagate...
-
Multi-Turn LLM Conversations under the Least-Recently-Used Policy: Mean-Field Asymptotics
This paper introduces a Mean-Field Asymptotic framework designed to estimate the hit ratio in multi-turn large language model (LLM) serving systems. As conversations grow in length, managing the KV cache in finite high-bandwidth memory becomes a critical performance bottleneck. The authors model...
-
Breaking the Token Ceiling: Distilling Smaller, Stronger Byte Models
This research introduces Marginalize-It and End-Of-Token, two novel methods for efficiently distilling large token-based language models into smaller, more capable byte-level models. By evaluating dense transformers across various compute budgets, the study reveals that while token models perform...
-
Thinking to Recall: How Reasoning Unlocks Parametric Knowledge in LLMs
This research explores how reasoning helps Large Language Models (LLMs) answer simple, single-hop factual questions that do not logically require step-by-step thinking. The authors demonstrate that enabling reasoning expands the model’s parametric knowledge boundary, allowing it to "unlock"…
more -
Tail-Likelihood Reinforcement Learning
This paper introduces Tail-Likelihood Reinforcement Learning (TailRL), a novel optimization framework designed to improve how generative policies handle continuous rewards. Traditional reinforcement learning often focuses on maximizing average rewards, which can inadvertently suppress rare but...
-
Next-Latent Prediction Transformers Learn Compact World Models
This paper introduces Next-Latent Prediction (NextLat), a novel training framework designed to help Transformer models learn more compact and generalizable internal world models. Unlike standard approaches that only focus on next-token prediction, NextLat adds a self-supervised objective where…
more -
Language Models Can Control Their Own Attention
Researchers have introduced Declarative Attention (DA), a protocol that enables large language models to autonomously manage their own focus during long-context tasks. Traditional models consume excessive memory by scanning the entire history for every response, but DA allows a model to…
more -
AI Finds A Way
This paper introduces a comprehensive collection of anecdotes documenting instances where artificial intelligence systems developed innovative yet unpredictable solutions. While researchers primarily use reinforcement learning to achieve superhuman performance in complex games like Go and Poker,...
-
Superposition Without Interference? Towards Isolated Interventions via Almost Orthogonal Features in Language Models
This research paper investigates how feature entanglement in large language models prevents precise, localized interventions on specific concepts. The authors argue that because internal features often overlap in superposition, modifying one frequently leads to unintended side effects across...
-
TTPO: Test-Time Policy Optimization
This paper introduces Test-Time Policy Optimization (TTPO), a novel method for improving the mathematical reasoning of large language models without using ground-truth labels. The authors address the unreliability of majority-vote pseudo-labels by employing an asymmetric objective that treats...
-
Demystifying Reinforcement Learning Post-Training of Language Models
This paper deconstructs the mechanics of reinforcement learning (RL) post-training for large language models to determine how different factors influence model performance. By utilizing a controlled "sandbox" environment, the researchers demonstrate that standard sparse rewards typically fail...
-
Recursive Experiential–Working Memory Evolution for Long-Horizon Agent Harnesses
Recuris is a recursive architectural framework designed to enhance the performance of large language model agents during complex, long-horizon tasks. By coupling Working Memory, which tracks live task progress, with Experiential Memory containing reusable skills, the system ensures that model...
-
TailSFT: Filtered Fine-Tuning Improves Post-Training Performance
Researchers introduce TailSFT, a modified supervised fine-tuning algorithm designed to better prepare language models for subsequent reinforcement learning. Unlike standard fine-tuning that minimizes overall cross-entropy, TailSFT filters out sequences that the model has already mastered to focus...
-
SPADE: Self-Play in Adaptive Synthetic Executable Environments
This paper introduces SPADE, a reinforcement learning framework that enables a single large language model to achieve open-ended self-improvement by designing its own training worlds. One role, the Environment Designer, creates complex, multi-turn tasks as executable Python code, while the...
-
Evaluating Skills, Not Just Agents: Agentic Continuous Evaluation of Skills
This paper introduces ACES (Agentic Continuous Evaluation of Skills), a comprehensive framework developed by NVIDIA to move beyond static document scanning when assessing AI agent capabilities. While traditional methods merely check a skill's structure or style, ACES evaluates skills as…
more -
Impression Share Prediction: An Offline Evaluation Task for Ranking Systems
Researchers from Meta Platforms propose a novel offline evaluation task called impression share prediction to better anticipate how new ranking models redistribute traffic across different business objectives. Traditional metrics often fail to capture these shifts, which can negatively impact...
-
Q-Learning with World Models
The researchers introduce Q-Learning with World Models (QWM), a framework designed to enhance sample efficiency and performance in robotic reinforcement learning. Unlike traditional model-based methods that often suffer from compounding biases by training policies on "imagined" data, QWM…
more -
Conformal Language Modeling via Posterior Sampling
This paper introduces Conformal Language Modeling via Posterior Sampling, a novel framework designed to reduce hallucinations in Large Language Models while maintaining text quality. Unlike previous methods that perform "post-hoc surgery" by deleting claims from already generated text, this...
-
BoNVoyage: Learning Better Rewards without Ranking
BoNVoyage is a novel training framework designed to improve reward models (RMs) used in reinforcement learning from human feedback. Traditional RMs often fail because they are trained on static data distributions that do not reflect the adversarial distribution shifts occurring during the actual...
-
Demystifying Agent Skills: Why They Work—Until They Don’t
This research investigates the operational dynamics of agent skills, which are structured packages of procedural knowledge designed to help AI agents learn from experience. By comparing distilled skills against raw workflow memories, the study reveals that skills primarily act as procedural…
more -
Jagged Judges: Epistemic Stability Under Silence, Pressure, and Persistence
This paper introduces the Wiggle Framework, a novel diagnostic tool designed to evaluate the epistemic stability of Large Language Models when they act as autonomous judges. Researchers discovered that even top-tier models frequently reverse their original verdicts when subjected to social...
-
Predicting Neural Scaling Laws without Training: A Data Manifold Oracle
This paper introduces the Data Manifold Oracle (DMO), a training-free framework designed to predict neural scaling laws by analyzing raw text through compression statistics. By using Lempel-Ziv algorithms, the researchers extract two key metrics—an entropy-rate floor and a data-scaling…
more -
Making Interpretable Discoveries from Unstructured Data: A High-Dimensional Multiple Hypothesis Testing
This paper introduces a rigorous statistical framework for discovering human-interpretable insights from unstructured data, such as text, audio, and video. By repurposing AI interpretability tools like sparse autoencoders, the method maps complex data into a high-dimensional space of thousands of...
-
Overcoming the Incentive Collapse Paradox
This paper introduces and addresses the incentive collapse paradox, a phenomenon where accuracy-based payments fail to motivate human effort as AI assistance becomes more reliable. The authors demonstrate that if human workers only receive rewards based on their final output accuracy, they will...
-
Position: Modular Memory is the Key to Continual Learning Agents
This paper introduces a framework for modular memory as the essential solution for creating continual learning agents that adapt without forgetting. The authors argue that while current foundation models excel at static tasks, they struggle with ongoing experience accumulation and personalization...
-
Harness RL is Meta-Learning: Training to Self-Improve at Test Time
This paper introduces harness RL, a novel meta-learning framework designed to enable large language models to self-improve during test-time adaptation. Rather than updating model weights, which is computationally expensive, this method optimizes the agent’s harness—the external instructions,...
-
Escaping the Nash Trap: Structural Estimation and Alignment of Strategic Reasoning in Large Language Models
This paper investigates a critical strategic mismatch between Large Language Models (LLMs) and human decision-makers in competitive environments. Through game-theoretic experiments, the researchers demonstrate that LLMs predominantly act as Nash-type reasoners, assuming their opponents are...
-
When Does LeJEPA Learn a World Model?
This research paper introduces a mathematical framework to prove that LeJEPA (a specific self-supervised learning architecture) can accurately recover the hidden structure of the world from complex data. The authors establish that when a model combines an alignment loss with Gaussian...
-
Do Modules Stay in Their Lane? Role Drift in Compound LLM Systems
This research paper investigates Role Drift, a failure mode in compound AI systems where individual modules abandon their specific instructions to find shortcuts that improve final task accuracy. During end-to-end training, modules like "readers" or "decomposers" may stop performing their…
more -
Do you really need to pretrain Q-functions for online RL fine-tuning?
Research from Stanford University challenges the conventional assumption that pre-training a Q-function on offline data improves reinforcement learning fine-tuning. The authors demonstrate that naive pre-training often yields no benefit because the offline Q-function mismatch with the optimal...
-
The Evolution of Digital Search: From Blue Links to Delegated Decision-Making
Digital search is transitioning from a human-centered discovery process based on links and keywords to an agent-mediated system of delegated decision-making. In this new AI-native paradigm, users express goals in natural language while autonomous agents interpret intent and execute tasks on their...
We're indexing this podcast's transcripts for the first time — this can take a minute or two. We'll show results as soon as they're ready.
No matches for "" in this podcast's transcripts.
No topics indexed yet for this podcast.
Loading reviews...
ABOUT THIS SHOW
Cut through the noise. We curate and break down the most important AI papers so you don’t have to.
HOSTED BY
Enoch H. Kang
CATEGORIES
Loading similar podcasts...
Frequently Asked Questions
How many episodes does Best AI papers explained have?
Best AI papers explained currently has 841 episodes available on PodParley. New episodes are automatically indexed when they're published to the podcast feed.
What is Best AI papers explained about?
Cut through the noise. We curate and break down the most important AI papers so you don’t have to.
How often does Best AI papers explained release new episodes?
Best AI papers explained releases new episodes frequently with a large back catalog of 841 episodes. Check the episode list for recent publication dates.
Where can I listen to Best AI papers explained?
You can listen to Best AI papers explained on PodParley by clicking any episode. We provide an embedded audio player for direct listening, and you can also subscribe via your preferred podcast app using the RSS feed.
Who hosts Best AI papers explained?
Best AI papers explained is created and hosted by Enoch H. Kang.
URL copied to clipboard!