Deep Papers
Deep Papers is a podcast series that explores important AI papers and research. Hosted by Arize AI founders and engineers, each episode profiles the people and techniques behind breakthroughs in machine learning.
Episodes
CUGA Agent: From Benchmarks to Business Impact of IBM's Generalist Agent
We dive into the latest paper from a team of researchers at IBM: "From Benchmarks to Business Impact: Deploying IBM Generalist Agent in Enterprise Production." We're excited to host several of the paper's authors, who walk us through the research and its implications. The paper reports IBM’s experience developing and piloting the Computer Using Generalist Agent (CUGA), which ha
TUMIX: Multi-Agent Test-Time Scaling with Tool-Use Mixture
We dive into the latest paper from Google and a team of academic researchers: "TUMIX: Multi-Agent Test-Time Scaling with Tool-Use Mixture."Hear from one of the paper's authors — Yongchao Chen, Research Scientist — walks through the research and its implications. The paper proposes Tool-Use Mixture (TUMIX), an ensemble framework that runs multiple agents in parallel, each employing d
Meta AI Researcher Explains ARE and Gaia2: Scaling Up Agent Environments and Evaluations
In our latest paper reading, we had the pleasure of hosting Grégoire Mialon — Research Scientist at Meta Superintelligence Labs — to walk us through Meta AI’s groundbreaking paper titled “ARE: scaling up agent environments and evaluations" and the new ARE and Gaia2 frameworks.Learn more about AI observability and evaluation, join the Arize AI Slack community or get the latest on LinkedIn and
Georgia Tech's Santosh Vempala Explains Why Language Models Hallucinate, His Research With OpenAI
Santosh Vempala, Frederick Storey II Chair of Computing and Distinguished Professor in the School of Computer Science at Georgia Tech, explains his paper co-authored by OpenAI's Adam Tauman Kalai, Ofir Nachum, and Edwin Zhang. Read the paper: Sign up for future AI research paper readings and author office hours. See LLM hallucination examples here for context.Learn more about AI observability
Atropos Health’s Arjun Mukerji, PhD, Explains RWESummary: A Framework and Test for Choosing LLMs to Summarize Real-World Evidence (RWE) Studies
Large language models are increasingly used to turn complex study output into plain-English summaries. But how do we know which models are safest and most reliable for healthcare? In this most recent community AI research paper reading, Arjun Mukerji, PhD – Staff Data Scientist at Atropos Health – walks us through RWESummary, a new benchmark designed to evaluate LLMs on summarizing real-world evid
Stan Miasnikov, Distinguished Engineer, AI/ML Architecture, Consumer Experience at Verizon Walks Us Through His New Paper
This episode dives into "Category-Theoretic Analysis of Inter-Agent Communication and Mutual Understanding Metric in Recursive Consciousness." The paper presents an extension of the Recursive Consciousness framework to analyze communication between agents and the inevitable loss of meaning in translation. We're thrilled to feature the paper's author, Stan Miasnikov, Distinguish
Small Language Models are the Future of Agentic AI
We had the privilege of hosting Peter Belcak – an AI Researcher working on the reliability and efficiency of agentic systems at NVIDIA – who walked us through his new paper making the rounds in AI circles titled “Small Language Models are the Future of Agentic AI.”The paper posits that small language models (SLMs) are sufficiently powerful, inherently more suitable, and necessarily more economical
Watermarking for LLMs and Image Models
In this AI research paper reading, we dive into "A Watermark for Large Language Models" with the paper's author John Kirchenbauer. This paper is a timely exploration of techniques for embedding invisible but detectable signals in AI-generated text. These watermarking strategies aim to help mitigate misuse of large language models by making machine-generated content distinguishable f
Self-Adapting Language Models: Paper Authors Discuss Implications
The authors of the new paper *Self-Adapting Language Models (SEAL)* shared a behind-the-scenes look at their work, motivations, results, and future directions.The paper introduces a novel method for enabling large language models (LLMs) to adapt their own weights using self-generated data and training directives — “self-edits.”Learn more about the Self-Adapting Language Models paper.Learn more abo
The Illusion of Thinking: What the Apple AI Paper Says About LLM Reasoning
This week we discuss The Illusion of Thinking, a new paper from researchers at Apple that challenges today’s evaluation methods and introduces a new benchmark: synthetic puzzles with controllable complexity and clean logic. Their findings? Large Reasoning Models (LRMs) show surprising failure modes, including a complete collapse on high-complexity tasks and a decline in reasoning effort as problem
Accurate KV Cache Quantization with Outlier Tokens Tracing
We discuss Accurate KV Cache Quantization with Outlier Tokens Tracing, a deep dive into improving the efficiency of LLM inference. The authors enhance KV Cache quantization, a technique for reducing memory and compute costs during inference, by introducing a method to identify and exclude outlier tokens that hurt quantization accuracy, striking a better balance between efficiency and performance.R
Scalable Chain of Thoughts via Elastic Reasoning
In this week's episode, we talk about Elastic Reasoning, a novel framework designed to enhance the efficiency and scalability of large reasoning models by explicitly separating the reasoning process into two distinct phases: thinking and solution. This separation allows for independent allocation of computational budgets, addressing challenges related to uncontrolled output lengths in real-wo
Sleep-time Compute: Beyond Inference Scaling at Test-time
What if your LLM could think ahead—preparing answers before questions are even asked?In this week's paper read, we dive into a groundbreaking new paper from researchers at Letta, introducing sleep-time compute: a novel technique that lets models do their heavy lifting offline, well before the user query arrives. By predicting likely questions and precomputing key reasoning steps, sleep-time c
LibreEval: The Largest Open Source Benchmark for RAG Hallucination Detection
For this week's paper read, we dive into our own research.We wanted to create a replicable, evolving dataset that can keep pace with model training so that you always know you're testing with data your model has never seen before. We also saw the prohibitively high cost of running LLM evals at scale, and have used our data to fine-tune a series of SLMs that perform just as well as their
AI Benchmark Deep Dive: Gemini 2.5 and Humanity's Last Exam
This week we talk about modern AI benchmarks, taking a close look at Google's recent Gemini 2.5 release and its performance on key evaluations, notably Humanity's Last Exam (HLE). In the session we covered Gemini 2.5's architecture, its advancements in reasoning and multimodality, and its impressive context window. We also talked about how benchmarks like HLE and ARC AGI 2 help us
Model Context Protocol (MCP)
We cover Anthropic’s groundbreaking Model Context Protocol (MCP). Though it was released in November 2024, we've been seeing a lot of hype around it lately, and thought it was well worth digging into. Learn how this open standard is revolutionizing AI by enabling seamless integration between LLMs and external data sources, fundamentally transforming them into capable, context-aware agents. We
AI Roundup: DeepSeek’s Big Moves, Claude 3.7, and the Latest Breakthroughs
This week, we're mixing things up a little bit. Instead of diving deep into a single research paper, we cover the biggest AI developments from the past few weeks.We break down key announcements, including:DeepSeek’s Big Launch Week: A look at FlashMLA (DeepSeek’s new approach to efficient inference) and DeepEP (their enhanced pretraining method).Claude 3.7 & Claude Code: What’s new with A
How DeepSeek is Pushing the Boundaries of AI Development
This week, we dive into DeepSeek. SallyAnn DeLucia, Product Manager at Arize, and Nick Luzio, a Solutions Engineer, break down key insights on a model that have dominating headlines for its significant breakthrough in inference speed over other models. What’s next for AI (and open source)? From training strategies to real-world performance, here’s what you need to know.Read our analysis of DeepSee
Multiagent Finetuning: A Conversation with Researcher Yilun Du
We talk to Google DeepMind Senior Research Scientist (and incoming Assistant Professor at Harvard), Yilun Du, about his latest paper, "Multiagent Finetuning: Self Improvement with Diverse Reasoning Chains." This paper introduces a multiagent finetuning framework that enhances the performance and diversity of language models by employing a society of agents with distinct roles, improving
Training Large Language Models to Reason in Continuous Latent Space
LLMs have typically been restricted to reason in the "language space," where chain-of-thought (CoT) is used to solve complex reasoning problems. But a new paper argues that language space may not always be the best for reasoning. In this paper read, we cover an exciting new technique from a team at Meta called Chain of Continuous Thought—also known as "Coconut." In the paper, &
LLMs as Judges: A Comprehensive Survey on LLM-Based Evaluation Methods
We discuss a major survey of work and research on LLM-as-Judge from the last few years. "LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods" systematically examines the LLMs-as-Judge framework across five dimensions: functionality, methodology, applications, meta-evaluation, and limitations. This survey gives us a birds eye view of the advantages, limitations and meth
Merge, Ensemble, and Cooperate! A Survey on Collaborative LLM Strategies
LLMs have revolutionized natural language processing, showcasing remarkable versatility and capabilities. But individual LLMs often exhibit distinct strengths and weaknesses, influenced by differences in their training corpora. This diversity poses a challenge: how can we maximize the efficiency and utility of LLMs?A new paper, "Merge, Ensemble, and Cooperate: A Survey on Collaborative Strate
Agent-as-a-Judge: Evaluate Agents with Agents
This week, we break down the “Agent-as-a-Judge” framework—a new agent evaluation paradigm that’s kind of like getting robots to grade each other’s homework. Where typical evaluation methods focus solely on outcomes or demand extensive manual work, this approach uses agent systems to evaluate agent systems, offering intermediate feedback throughout the task-solving process. With the power to unlock
Introduction to OpenAI's Realtime API
We break down OpenAI’s realtime API. Learn how to seamlessly integrate powerful language models into your applications for instant, context-aware responses that drive user engagement. Whether you’re building chatbots, dynamic content tools, or enhancing real-time collaboration, we walk through the API’s capabilities, potential use cases, and best practices for implementation. Learn more about AI o
Swarm: OpenAI's Experimental Approach to Multi-Agent Systems
As multi-agent systems grow in importance for fields ranging from customer support to autonomous decision-making, OpenAI has introduced Swarm, an experimental framework that simplifies the process of building and managing these systems. Swarm, a lightweight Python library, is designed for educational purposes, stripping away complex abstractions to reveal the foundational concepts of multi-agent a
KV Cache Explained
In this episode, we dive into the intriguing mechanics behind why chat experiences with models like GPT often start slow but then rapidly pick up speed. The key? The KV cache. This essential but under-discussed component enables the seamless and snappy interactions we expect from modern AI systems.Harrison Chu breaks down how the KV cache works, how it relates to the transformer architecture, and
The Shrek Sampler: How Entropy-Based Sampling is Revolutionizing LLMs
In this byte-sized podcast, Harrison Chu, Director of Engineering at Arize, breaks down the Shrek Sampler. This innovative Entropy-Based Sampling technique--nicknamed the 'Shrek Sampler--is transforming LLMs. Harrison talks about how this method improves upon traditional sampling strategies by leveraging entropy and varentropy to produce more dynamic and intelligent responses. Explore its pot
Google's NotebookLM and the Future of AI-Generated Audio
This week, Aman Khan and Harrison Chu explore NotebookLM’s unique features, including its ability to generate realistic-sounding podcast episodes from text (but this podcast is very real!). They dive into some technical underpinnings of the product, specifically the SoundStorm model used for generating high-quality audio, and how it leverages a hierarchical vector quantization approach (RVQ) to ma
Exploring OpenAI's o1-preview and o1-mini
OpenAI recently released its o1-preview, which they claim outperforms GPT-4o on a number of benchmarks. These models are designed to think more before answering and handle complex tasks better than their other models, especially science and math questions. We take a closer look at their latest crop of o1 models, and we also highlight some research our team did to see how they stack up against Clau
Breaking Down Reflection Tuning: Enhancing LLM Performance with Self-Learning
A recent announcement on X boasted a tuned model with pretty outstanding performance, and claimed these results were achieved through Reflection Tuning. However, people were unable to reproduce the results. We dive into some recent drama in the AI community as a jumping off point for a discussion about Reflection 70B.In 2023, there was a paper written about Reflection Tuning that this new model (R
Composable Interventions for Language Models
This week, we're excited to be joined by Kyle O'Brien, Applied Scientist at Microsoft, to discuss his most recent paper, Composable Interventions for Language Models. Kyle and his team present a new framework, composable interventions, that allows for the study of multiple interventions applied sequentially to the same language model. The discussion will cover their key findings from ext
Judging the Judges: Evaluating Alignment and Vulnerabilities in LLMs-as-Judges
This week’s paper presents a comprehensive study of the performance of various LLMs acting as judges. The researchers leverage TriviaQA as a benchmark for assessing objective knowledge reasoning of LLMs and evaluate them alongside human annotations which they find to have a high inter-annotator agreement. The study includes nine judge models and nine exam-taker models – both base and instruction-t
Breaking Down Meta's Llama 3 Herd of Models
Meta just released Llama 3.1 405B–according to them, it’s “the first openly available model that rivals the top AI models when it comes to state-of-the-art capabilities in general knowledge, steerability, math, tool use, and multilingual translation.” Will the latest Llama herd ignite new applications and modeling paradigms like synthetic data generation? Will it enable the improvement and trainin
DSPy Assertions: Computational Constraints for Self-Refining Language Model Pipelines
Chaining language model (LM) calls as composable modules is fueling a new way of programming, but ensuring LMs adhere to important constraints requires heuristic “prompt engineering.” The paper this week introduces LM Assertions, a programming construct for expressing computational constraints that LMs should satisfy. The researchers integrated their constructs into the recent DSPy programming mod
RAFT: Adapting Language Model to Domain Specific RAG
Where adapting LLMs to specialized domains is essential (e.g., recent news, enterprise private documents), we discuss a paper that asks how we adapt pre-trained LLMs for RAG in specialized domains. SallyAnn DeLucia is joined by Sai Kolasani, researcher at UC Berkeley’s RISE Lab (and Arize AI Intern), to talk about his work on RAFT: Adapting Language Model to Domain Specific RAG. RAFT (Retrieval-Au
LLM Interpretability and Sparse Autoencoders: Research from OpenAI and Anthropic
It’s been an exciting couple weeks for GenAI! Join us as we discuss the latest research from OpenAI and Anthropic. We’re excited to chat about this significant step forward in understanding how LLMs work and the implications it has for deeper understanding of the neural activity of language models. We take a closer look at some recent research from both OpenAI and Anthropic. These two recent paper
Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models' Alignment
We break down the paper--Trustworthy LLMs: A Survey and Guideline for Evaluating Large Language Models' Alignment.Ensuring alignment (aka: making models behave in accordance with human intentions) has become a critical task before deploying LLMs in real-world applications. However, a major challenge faced by practitioners is the lack of clear guidance on evaluating whether LLM outputs align w
Breaking Down EvalGen: Who Validates the Validators?
Due to the cumbersome nature of human evaluation and limitations of code-based evaluation, Large Language Models (LLMs) are increasingly being used to assist humans in evaluating LLM outputs. Yet LLM-generated evaluators often inherit the problems of the LLMs they evaluate, requiring further human validation.This week’s paper explores EvalGen, a mixed-initative approach to aligning LLM-generated e
Keys To Understanding ReAct: Synergizing Reasoning and Acting in Language Models
This week we explore ReAct, an approach that enhances the reasoning and decision-making capabilities of LLMs by combining step-by-step reasoning with the ability to take actions and gather information from external sources in a unified framework.Learn more about AI observability and evaluation, join the Arize AI Slack community or get the latest on LinkedIn and X.
Demystifying Chronos: Learning the Language of Time Series
This week, we’ve covering Amazon’s time series model: Chronos. Developing accurate machine-learning-based forecasting models has traditionally required substantial dataset-specific tuning and model customization. Chronos however, is built on a language model architecture and trained with billions of tokenized time series observations, enabling it to provide accurate zero-shot forecasts matching or
Anthropic Claude 3
This week we dive into the latest buzz in the AI world – the arrival of Claude 3. Claude 3 is the newest family of models in the LLM space, and Opus Claude 3 ( Anthropic's "most intelligent" Claude model ) challenges the likes of GPT-4.The Claude 3 family of models, according to Anthropic "sets new industry benchmarks," and includes "three state-of-the-art models in a
Reinforcement Learning in the Era of LLMs
We’re exploring Reinforcement Learning in the Era of LLMs this week with Claire Longo, Arize’s Head of Customer Success. Recent advancements in Large Language Models (LLMs) have garnered wide attention and led to successful products such as ChatGPT and GPT-4. Their proficiency in adhering to instructions and delivering harmless, helpful, and honest (3H) responses can largely be attributed to the t
Sora: OpenAI’s Text-to-Video Generation Model
This week, we discuss the implications of Text-to-Video Generation and speculate as to the possibilities (and limitations) of this incredible technology with some hot takes. Dat Ngo, ML Solutions Engineer at Arize, is joined by community member and AI Engineer Vibhu Sapra to review OpenAI’s technical report on their Text-To-Video Generation Model: Sora.According to OpenAI, “Sora can generate video
RAG vs Fine-Tuning
This week, we’re discussing "RAG vs Fine-Tuning: Pipelines, Tradeoff, and a Case Study on Agriculture." This paper explores a pipeline for fine-tuning and RAG, and presents the tradeoffs of both for multiple popular LLMs, including Llama2-13B, GPT-3.5, and GPT-4. The authors propose a pipeline that consists of multiple stages, including extracting information from PDFs, generating questi
Phi-2 Model
We dive into Phi-2 and some of the major differences and use cases for a small language model (SLM) versus an LLM.With only 2.7 billion parameters, Phi-2 surpasses the performance of Mistral and Llama-2 models at 7B and 13B parameters on various aggregated benchmarks. Notably, it achieves better performance compared to 25x larger Llama-2-70B model on multi-step reasoning tasks, i.e., coding and ma
HyDE: Precise Zero-Shot Dense Retrieval without Relevance Labels
We discuss HyDE: a thrilling zero-shot learning technique that combines GPT-3’s language understanding with contrastive text encoders. HyDE revolutionizes information retrieval and grounding in real-world data by generating hypothetical documents from queries and retrieving similar real-world documents. It outperforms traditional unsupervised retrievers, rivaling fine-tuned retrievers across diver
A Deep Dive Into Generative's Newest Models: Gemini vs Mistral (Mixtral-8x7B)–Part I
For the last paper read of the year, Arize CPO & Co-Founder, Aparna Dhinakaran, is joined by a Dat Ngo (ML Solutions Architect) and Aman Khan (Product Manager) for an exploration of the new kids on the block: Gemini and Mixtral-8x7B. There's a lot to cover, so this week's paper read is Part I in a series about Mixtral and Gemini. In Part I, we provide some background and context for
How to Prompt LLMs for Text-to-SQL: A Study in Zero-shot, Single-domain, and Cross-domain Settings
We’re thrilled to be joined by Shuaichen Chang, LLM researcher and the author of this week’s paper to discuss his findings. Shuaichen’s research investigates the impact of prompt constructions on the performance of large language models (LLMs) in the text-to-SQL task, particularly focusing on zero-shot, single-domain, and cross-domain settings. Shuaichen and his team explore various strategies for
The Geometry of Truth: Emergent Linear Structure in LLM Representation of True/False Datasets
For this paper read, we’re joined by Samuel Marks, Postdoctoral Research Associate at Northeastern University, to discuss his paper, “The Geometry of Truth: Emergent Linear Structure in LLM Representation of True/False Datasets.” Samuel and his team curated high-quality datasets of true/false statements and used them to study in detail the structure of LLM representations of truth. Overall, they p
Towards Monosemanticity: Decomposing Language Models With Dictionary Learning
In this paper read, we discuss “Towards Monosemanticity: Decomposing Language Models Into Understandable Components,” a paper from Anthropic that addresses the challenge of understanding the inner workings of neural networks, drawing parallels with the complexity of human brain function. It explores the concept of “features,” (patterns of neuron activations) providing a more interpretable way to d
RankVicuna: Zero-Shot Listwise Document Reranking with Open-Source Large Language Models
We discuss RankVicuna, the first fully open-source LLM capable of performing high-quality listwise reranking in a zero-shot setting. While researchers have successfully applied LLMs such as ChatGPT to reranking in an information retrieval context, such work has mostly been built on proprietary models hidden behind opaque API endpoints. This approach yields experimental results that are not reprodu
Explaining Grokking Through Circuit Efficiency
Join Arize Co-Founder & CEO Jason Lopatecki, and ML Solutions Engineer, Sally-Ann DeLucia, as they discuss “Explaining Grokking Through Circuit Efficiency." This paper explores novel predictions about grokking, providing significant evidence in favor of its explanation. Most strikingly, the research conducted in this paper demonstrates two novel and surprising behaviors: ungrokking, in wh
Large Content And Behavior Models To Understand, Simulate, And Optimize Content And Behavior
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. In this episode, we discuss the paper, “Large Content And Behavior Models To Understand, Simulate, And Optimize Content And Behavior.” This episode is led by SallyAnn Delucia (ML Solutions Engineer, A
Skeleton of Thought: LLMs Can Do Parallel Decoding
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. In this paper reading, we explore the paper ‘Skeleton-of-Thought’ (SoT) approach, aimed at reducing large language model latency while enhancing answer quality. This episode is led by Aparna Dhinakaran
Llama 2: Open Foundation and Fine-Tuned Chat Models
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. This episode is led by Aparna Dhinakaran ( Chief Product Officer, Arize AI) and Michael Schiff (Chief Technology Officer, Arize AI), as they discuss the paper "Llama 2: Open Foundation and Fine-Tu
Lost in the Middle: How Language Models Use Long Contexts
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. This episode is led by Sally-Ann DeLucia and Amber Roberts, as they discuss the paper "Lost in the Middle: How Language Models Use Long Contexts." This paper examines how well language models
Orca: Progressive Learning from Complex Explanation Traces of GPT-4
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Hosted by AI Pub creator Brian Burns and Arize AI founders Jason Lopatecki and Aparna Dhinakaran, each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning.In this episode, we talk about Orca. Recent research focuses on improving smaller models through imitati
Toolformer: Training LLMs To Use Tools
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Hosted by AI Pub creator Brian Burns and Arize AI founders Jason Lopatecki and Aparna Dhinakaran, each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. In this episode, we interview Timo Schick and Thomas Scialom, the Research Scientists at Meta AI behind
Hungry Hungry Hippos - H3
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Hosted by AI Pub creator Brian Burns and Arize AI founders Jason Lopatecki and Aparna Dhinakaran, each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. In this episode, we interview Dan Fu and Tri Dao, inventors of "Hungry Hungry Hippos" (aka &q
ChatGPT and InstructGPT: Aligning Language Models to Human Intention
Deep Papers is a podcast series featuring deep dives on today’s seminal AI papers and research. Hosted by AI Pub creator Brian Burns and Arize AI founders Jason Lopatecki and Aparna Dhinakaran, each episode profiles the people and techniques behind cutting-edge breakthroughs in machine learning. In this first episode, we’re joined by Long Ouyang and Ryan Lowe, research scientists at OpenAI and cre
