AI Post Transformers

The January 30, 2025 paper introduces LLM-AutoDiff, a novel framework for Automatic Prompt Engineering (APE) that allows for the optimization of complex Large Language Model (LLM) workflows. This framework models an entire LLM application—including multiple LLM calls, functional components like retrievers, and cyclical operations—as a directed, auto-differentiable graph. By treating textual inputs as trainable parameters, LLM-AutoDiff uses a separate "backward engine" LLM to generate textual gradients (feedback) that guide an optimizer LLM to revise prompts, effectively automating the manual and labor-intensive process of prompt engineering. The paper details several technical advances, such as pass-through gradients for functional nodes and time-sequential gradients for cyclic structures, to ensure accurate error attribution across multi-component pipelines, ultimately demonstrating improved accuracy and efficiency over existing textual gradient and few-shot baselines. Source: January 30, 2025 LLM-AutoDiff: Auto-Differentiate Any LLM Workflow https://arxiv.org/pdf/2501.16673

The May 20, 2024 academic paper explores the metacognitive capabilities of Large Language Models (LLMs), specifically focusing on mathematical problem-solving. The core approach involves developing a method for a powerful LLM, such as GPT-4, to identify and label mathematical questions with specific skills, which are then organized into broader, interpretable categories. This process creates a Skill Exemplar Repository containing skill names matched with question-answer pairs. Experiments validate that providing an LLM with these skill labels and associated examples as in-context prompts significantly improves accuracy on challenging math datasets like MATH and GSM8K, outperforming baseline prompting techniques like Chain-of-Thought. Furthermore, the skill knowledge transferred effectively to other, less powerful LLMs and different math datasets, demonstrating the utility of this LLM-generated metacognitive framework. Source: https://arxiv.org/pdf/2405.12205

We provide a review of the evolution of value of Page Rank to Random Walk with Random Restart and it's application to neural networks focusing on five research papers dating from the original page rank to 2025. They collectively focus on methods for learning on graphs, particularly through the use of Random Walk Neural Networks (RWNNs) and related random walk algorithms. One primary source introduces RWNNs, detailing their architecture, which involves a random walk generating a machine-readable record processed by a deep neural network, demonstrating that these models can achieve universal approximation of graph functions and overcome issues like over-smoothing found in Message Passing Neural Networks (MPNNs). This source also explores techniques like anonymization and named neighbors for walk recording and includes experimental results on graph isomorphism and transductive classification using language models like DeBERTa and Llama 3. The other sources provide brief contextual support, mentioning Random Walk with Restart (RWR) parameters and evaluation criteria like Relative Accuracy and Relative Score for related graph applications and datasets, suggesting connections to established graph algorithms such as PageRank.Sources:2025:REVISITING RANDOM WALKS FOR LEARNING ON GRAPHShttps://proceedings.iclr.cc/paper_files/paper/2025/file/cd51b67dcb19db4e9f0022f500076b00-Paper-Conference.pdfOctober 3, 2022:Universal Multilayer Network Exploration byRandom Walk with Restarthttps://arxiv.org/pdf/2107.045652020:Random Walk Graph Neural Networkshttps://proceedings.neurips.cc/paper/2020/file/ba95d78a7c942571185308775a97a3a0-Paper.pdf2006:Fast Random Walk with Restart and Its Applicationshttps://www.cs.cmu.edu/~htong/pdf/ICDM06_tong.pdfJanuary 29, 1998:The Page Rank Citation Ranking: Bringing Order to the Webhttps://www.cis.upenn.edu/~mkearns/teaching/NetworkedLife/pagerank.pdf

We review two papers on Spectral Gap, one 2021 and another from 2025. The first source presents the Spectral Attention Network (SAN), a novel Transformer-based architecture for graph neural networks that addresses the difficulty of defining positional encodings in graphs by leveraging the full Laplacian spectrum to learn node positions. This approach, which involves a Learned Positional Encoding (LPE), enables the fully-connected Transformer to overcome limitations of traditional Graph Neural Networks (GNNs) like over-squashing and achieves competitive or superior performance on standard benchmarks. The second source analyzes the stability and signal propagation in standard softmax-based attention layers of Transformers at initialization, identifying that a spectral gap in the attention matrix causes rank collapse both in the width and depth of the network, which hinders effective information flow and leads to exploding gradients. To remedy this, the authors propose a simple modification that removes the dominant outlier eigenvalue, demonstrating that this fix significantly mitigates rank collapse and stabilizes gradient growth in deep Transformer models. Both sources focus on improving the theoretical foundations and performance of attention mechanisms, with the first applying Transformers to graphs using spectral theory and the second addressing intrinsic instability issues in the core Transformer architecture.Sources:October 27, 2021:Rethinking Graph Transformers with SpectralAttentionhttps://arxiv.org/pdf/2106.03893June 16, 2025:Mind the Gap: a Spectral Analysis of Rank Collapseand Signal Propagation in Attention Layershttps://arxiv.org/pdf/2410.07799

The April 24, 2025 academic paper introduces Tempo, a novel scheduling system designed to optimize Large Language Model (LLM) serving by addressing the wide variety of Service Level Objectives (SLOs) in modern LLM applications. The authors categorize requests into three types—latency-sensitive, throughput-intensive, and collective requests—each with distinct performance requirements that existing schedulers fail to manage effectively. Tempo maximizes "service gain" by allocating just enough serving bandwidth to meet each request’s SLO, utilizing a hybrid scheduling strategy that relies on lightweight prediction models for conservative initial estimates of response length and dependency-graph matching for complex workflows. Evaluations demonstrate that Tempo significantly outperforms state-of-the-art systems in terms of both service gain and SLO goodput across diverse workloads and models. Source: April 24, 2025 Tempo: Application-aware LLM Serving with Mixed SLO Requirements https://arxiv.org/pdf/2504.20068

The 2024 paper introduces SYMPHONY, a novel system designed to improve memory management and scheduling for Large Language Model (LLM) inference workloads, particularly those involving multi-turn interactions like chatbots and AI agents. The authors, researchers from the University of Texas-Austin and the University of Wisconsin-Madison, explain that existing LLM serving engines either waste computation by recomputing Key-Value (K,V) caches or suffer from load imbalance by offloading caches to host memory, creating stateful workloads. SYMPHONY addresses these issues by using "advisory requests"—signals indicating the likely arrival of a new request—to proactively migrate K,V caches off the critical serving path, thereby enabling fine-grained scheduling and load balancing. Evaluation results demonstrate that SYMPHONY significantly reduces latency and can handle over eight times the number of requests compared to state-of-the-art baselines. Source: December 21, 2024 SYMPHONY: Improving Memory Management for LLM Inference Workloads https://arxiv.org/pdf/2412.16434

The May 21, 2024 paper introduces Vidur, a new, high-fidelity simulation framework designed to optimize the deployment and performance of Large Language Model (LLM) inference. The authors explain that experimentally optimizing LLM deployment is prohibitively expensive, requiring exploration of a vast configuration space of system parameters like parallelization strategies and batching techniques, which can cost hundreds of thousands of dollars and thousands of GPU hours. Vidur addresses this by using predictive modeling and experimental profiling of LLM operators to estimate end-to-end performance metrics, achieving less than 9% error in latency estimation. Complementing the simulator is Vidur-Search, a configuration search tool that leverages Vidur to automatically identify the most cost-effective deployment settings that meet application performance constraints, reducing optimization time from months of GPU time to approximately one hour on a CPU machine. The research emphasizes that the optimal configuration depends on both the LLM and the specific workload trace, justifying the need for a rapid simulation tool like Vidur. Source: May 21, 2024 VIDUR: A LARGE-SCALE SIMULATION FRAMEWORK FOR LLM INFERENCE https://arxiv.org/pdf/2405.05465

The May 24, 2025 UT Austin paper introduces the Anchored Diffusion Language Model (ADLM), a novel approach that aims to improve discrete language modeling by addressing the limitations of traditional Autoregressive (AR) and standard Diffusion Language Models (DLMs). AR models, such as GPT-3, generate text sequentially and struggle with complex reasoning, while existing DLMs, which use iterative masked-token prediction, lag behind AR models in quality. ADLM enhances DLMs by incorporating anchor tokens—semantically important words—that guide the denoising process through a two-component architecture: an anchor network and a denoising network. The training is formalized through the Anchored Negative Evidence Lower Bound (ANELBO) objective, which encourages the anchor network to predict these key tokens early, significantly reducing the sample complexity and improving both likelihood modeling (achieving better perplexity scores) and generated text quality. Furthermore, the concept of anchoring is successfully extended to AR models via Anchored Chain-of-Thought (ACoT) fine-tuning, demonstrating improved performance on math and logical reasoning tasks. Source: https://arxiv.org/pdf/2505.18456

The October 29 2025 Google research paper introduces Supervised Reinforcement Learning (SRL), a novel framework designed to improve the complex, multi-step reasoning abilities of large language models (LLMs). The core issue addressed is that conventional training methods like Supervised Fine-Tuning (SFT) and outcome-based Reinforcement Learning with Verifiable Rewards (RLVR) struggle with difficult problems because they either overfit rigid expert paths or receive only sparse, uninformative final outcome rewards. SRL overcomes this by reformulating problem-solving as a sequence of logical "actions" and providing dense, step-wise rewards based on the similarity between the model's actions and expert demonstrations. Through extensive experiments, the paper demonstrates that SRL significantly outperforms baseline methods on challenging mathematical reasoning and software engineering benchmarks, especially when used to initialize training before subsequent refinement with RLVR. Source: https://arxiv.org/pdf/2510.25992

These two papers (years 2017, 2022) introduce and then apply the Gumbel-Softmax distribution as a differentiable gradient estimator for categorical and discrete latent variables in neural networks. The first paper, "Categorical Reparameterization with Gumbel-Softmax," proposes this distribution to address the challenge of backpropagating through non-differentiable sampling operations, demonstrating its effectiveness in tasks like structured output prediction and generative modeling, where it outperforms existing gradient estimators and allows for significant speedups in semi-supervised classification. The second paper, "Gumbel-Softmax Selective Networks," leverages this same reparameterization trick to train selective neural networks with an integrated, binary option to abstain from predicting when uncertain, thereby establishing an end-to-end differentiable framework for selective regression and classification tasks. Collectively, the sources present the Gumbel-Softmax technique as a general, principled method for enabling gradient flow through discrete or binary choices in neural network training.Sources:https://arxiv.org/pdf/1611.01144https://arxiv.org/pdf/2211.10564