Deep Learning Weekly: Issue 407
The State of Enterprise AI in 2025, Letta Leaderboard: Benchmarking LLMs on Agentic Memory, a paper on Insights into DeepSeek-V3, and many more!
This week in deep learning, we bring you The State of Enterprise AI in 2025: Measured Progress Over Hype, How Can AI Researchers Save Energy? By Going Backward., and a paper on Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures.
You may also enjoy QwenLong-L1, Just know stuff, proteinML edition, a paper on RenderFormer: Transformer-based Neural Rendering of Triangle Meshes with Global Illumination, and more!
As always, happy reading and hacking. If you have something you think should be in next week's issue, find us on Twitter: @dl_weekly.
Until next week!
Industry
QwenLong-L1 solves long-context reasoning challenge that stumps current LLMs
Alibaba Group has introduced QwenLong-L1, a new framework that enables large language models to reason over extremely long inputs.
The State of Enterprise AI in 2025: Measured Progress Over Hype
A survey-based blog post summarizing the state of enterprise AI adoption in 2025, highlighting the continued dominance of traditional search, prioritization of internal use cases, and key implementation challenges.
How Can AI Researchers Save Energy? By Going Backward.
An article about reversible computing as a potential energy-saving technology for AI, detailing its history, principles, challenges, and application in parallel processing.
ElevenLabs debuted Conversational AI 2.0, a significant upgrade to its platform for building advanced voice agents for enterprise use cases.
Prepared raises $80M to expand AI-powered emergency response platform
Prepared, a startup offering AI-powered solutions for emergency response, has raised $80 million in new funding.
MLOps & LLMOps
Letta Leaderboard: Benchmarking LLMs on Agentic Memory
The Letta team announced the Letta Leaderboard, a comprehensive benchmark suite that evaluates how effectively LLMs manage agentic memory.
RAG is dead, long live agentic retrieval
A blog post explaining the evolution from naive RAG to agentic retrieval strategies for LLMs, detailing various retrieval modes and techniques for building knowledge agents across multiple indices.
Accelerating Text-to-SQL Inference on Vanna with NVIDIA NIM for Faster Analytics
A tutorial that shows how to optimize Vanna’s text-to-SQL solution using NVIDIA NIM, accelerated inference microservices that deliver optimized endpoints for generative AI models.
Learning
SelfCheckGPT for LLM Evaluation
A blog post that explores a reference-free, zero-resource method as an alternative to LLM-as-a-judge for detecting hallucinations in language models.
RLHF 101: A Technical Tutorial on Reinforcement Learning from Human Feedback
A technical tutorial detailing the pipeline for Reinforcement Learning from Human Feedback (RLHF), covering data generation, reward model inference, filtering, tokenization, and training.
Just know stuff, proteinML edition
A comprehensive blog post providing a guide to protein machine learning for beginners, covering essential data types, deep learning models, computational methods, and fundamental biology concepts.
Libraries & Code
poloclub/transformer-explainer
An interactive visualization tool designed to help anyone learn how Transformer-based models like GPT work.
Papers & Publications
RenderFormer: Transformer-based Neural Rendering of Triangle Meshes with Global Illumination
Abstract:
We present RenderFormer, a neural rendering pipeline that directly renders an image from a triangle-based representation of a scene with full global illumination effects and that does not require per-scene training or fine-tuning. Instead of taking a physics-centric approach to rendering, we formulate rendering as a sequence-to-sequence transformation where a sequence of tokens representing triangles with reflectance properties is converted to a sequence of output tokens representing small patches of pixels. RenderFormer follows a two stage pipeline: a view-independent stage that models triangle-to-triangle light transport, and a view-dependent stage that transforms a token representing a bundle of rays to the corresponding pixel values guided by the triangle-sequence from the view-independent stage. Both stages are based on the transformer architecture and are learned with minimal prior constraints. We demonstrate and evaluate RenderFormer on scenes with varying complexity in shape and light transport.
Insights into DeepSeek-V3: Scaling Challenges and Reflections on Hardware for AI Architectures
Abstract:
The rapid scaling of large language models (LLMs) has unveiled critical limitations in current hardware architectures, including constraints in memory capacity, computational efficiency, and interconnection bandwidth. DeepSeek-V3, trained on 2,048 NVIDIA H800 GPUs, demonstrates how hardware-aware model co-design can effectively address these challenges, enabling cost-efficient training and inference at scale. This paper presents an in-depth analysis of the DeepSeek-V3/R1 model architecture and its AI infrastructure, highlighting key innovations such as Multi-head Latent Attention (MLA) for enhanced memory efficiency, Mixture of Experts (MoE) architectures for optimized computation-communication trade-offs, FP8 mixed-precision training to unlock the full potential of hardware capabilities, and a Multi-Plane Network Topology to minimize cluster-level network overhead. Building on the hardware bottlenecks encountered during DeepSeek-V3's development, we engage in a broader discussion with academic and industry peers on potential future hardware directions, including precise low-precision computation units, scale-up and scale-out convergence, and innovations in low-latency communication fabrics. These insights underscore the critical role of hardware and model co-design in meeting the escalating demands of AI workloads, offering a practical blueprint for innovation in next-generation AI systems.
 | A guest post by
|