Addition in Four Movements: Mapping Layer-wise Information Trajectories in LLMs

• URL: http://arxiv.org/abs/2506.07824v2
• Date: Tue, 09 Sep 2025 16:35:39 GMT
• Title: Addition in Four Movements: Mapping Layer-wise Information Trajectories in LLMs
• Abstract summary: We analyze the internal arithmetic processes in LLaMA-3-8B-Instruct.<n>Inspired by the step-by-step manner in which humans perform addition, we propose and analyze a coherent four-stage trajectory.<n>Near the output, the model organizes and generates the final content, with the correct token reliably occupying the top rank.
• Score: 0.4086920237000588
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Multi-digit addition is a clear probe of the computational power of large language models. To dissect the internal arithmetic processes in LLaMA-3-8B-Instruct, we combine linear probing with logit-lens inspection. Inspired by the step-by-step manner in which humans perform addition, we propose and analyze a coherent four-stage trajectory in the forward pass:Formula-structure representations become linearly decodable first, while the answer token is still far down the candidate list.Core computational features then emerge prominently.At deeper activation layers, numerical abstractions of the result become clearer, enabling near-perfect detection and decoding of the individual digits in the sum.Near the output, the model organizes and generates the final content, with the correct token reliably occupying the top rank.This trajectory suggests a hierarchical process that favors internal computation over rote memorization. We release our code and data to facilitate reproducibility.

Related papers

• Hidden Decoding at Scale: Latent Computation Scaling for Large Language Models [75.80275843320511]
  We study whether an existing backbone can keep improving by allocating more computation to each token while leaving the Transformer backbone fixed.<n>We propose Hidden Decoding, a sequence-length scaling method applied during continued pretraining.
  arXiv  Detail & Related papers  (2026-07-09T07:37:59Z)
• Reading Between the Dots: Decoding Hidden Computation across Filler Tokens [7.203846680259872]
  LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT)<n>This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning.<n>We introduce an unsupervised decoding pipeline that takes only hidden states as input and recovers intermediate values with 80-95% accuracy.
  arXiv  Detail & Related papers  (2026-07-03T17:18:34Z)
• Tracing Computation Density in LLMs [8.097942585959956]
  Transformer-based large language models (LLMs) are comprised of billions of parameters arranged in deep and wide computational graphs.<n>We introduce the s-Trace method to efficiently estimate the subgraph of size s that best approximates a full model output.
  arXiv  Detail & Related papers  (2026-05-26T13:55:08Z)
• From Early Encoding to Late Suppression: Interpreting LLMs on Character Counting Tasks [49.57538588967748]
  Large language models (LLMs) exhibit failures on elementary symbolic tasks such as character counting in a word, despite excelling on complex benchmarks.<n>We uncover a consistent phenomenon across modern architectures, including LLaMA, Qwen, and Gemma.<n>We show that symbolic reasoning failures in LLMs are not due to missing representations or insufficient scale, but arise from structured interference within the model's computation graph.
  arXiv  Detail & Related papers  (2026-04-01T11:40:12Z)
• DyLLM: Efficient Diffusion LLM Inference via Saliency-based Token Selection and Partial Attention [2.7422645382944935]
  We present DyLLM, a training-free inference framework that accelerates decoding by selectively computing only salient tokens.<n>Across diverse reasoning and code-generation benchmarks, DyLLM achieves up to 9.6x higher throughput.
  arXiv  Detail & Related papers  (2026-03-09T07:02:01Z)
• Step-Level Sparse Autoencoder for Reasoning Process Interpretation [48.99201531966593]
  Large Language Models (LLMs) have achieved strong complex reasoning capabilities through Chain-of-Thought (CoT) reasoning.<n>We propose step-level sparse autoencoder (SSAE), which serves as an analytical tool to disentangle different aspects of LLMs' reasoning steps into sparse features.<n> Experiments on multiple base models and reasoning tasks show the effectiveness of the extracted features.
  arXiv  Detail & Related papers  (2026-03-03T14:25:02Z)
• Learning on LLM Output Signatures for gray-box Behavior Analysis [52.81120759532526]
  Large Language Models (LLMs) have achieved widespread adoption, yet our understanding of their behavior remains limited.<n>We develop a transformer-based approach to process contamination and data detection in gray-box settings.<n>Our approach achieves superior performance on hallucination and data detection in gray-box settings, significantly outperforming existing baselines.
  arXiv  Detail & Related papers  (2025-03-18T09:04:37Z)
• Rethinking End-to-End 2D to 3D Scene Segmentation in Gaussian Splatting [86.15347226865826]
  We design a new end-to-end object-aware lifting approach, named Unified-Lift.<n>We augment each Gaussian point with an additional Gaussian-level feature learned using a contrastive loss to encode instance information.<n>We conduct experiments on three benchmarks: LERF-Masked, Replica, and Messy Rooms.
  arXiv  Detail & Related papers  (2025-03-18T08:42:23Z)
• Learning to Add, Multiply, and Execute Algorithmic Instructions Exactly with Neural Networks [5.3800094588915375]
  We study the training dynamics of two-layer fully connected networks in the infinite-width limit.<n>We show how a sufficiently large ensemble of such models can be trained to execute exactly, with high probability.<n>We show how this can be efficiently achieved using only logarithmically many training data.
  arXiv  Detail & Related papers  (2025-02-24T00:50:02Z)
• LASE: Learned Adjacency Spectral Embeddings [7.612218105739107]
  We learn nodal Adjacency Spectral Embeddings (ASE) from graph inputs.<n>LASE is interpretable, parameter efficient, robust to inputs with unobserved edges.<n>LASE layers combine Graph Convolutional Network (GCN) and fully-connected Graph Attention Network (GAT) modules.
  arXiv  Detail & Related papers  (2024-12-23T17:35:19Z)
• FIRP: Faster LLM inference via future intermediate representation prediction [54.897493351694195]
  FIRP generates multiple tokens instead of one at each decoding step. We conduct extensive experiments, showing a speedup ratio of 1.9x-3x in several models and datasets.
  arXiv  Detail & Related papers  (2024-10-27T15:53:49Z)
• COrAL: Order-Agnostic Language Modeling for Efficient Iterative Refinement [80.18490952057125]
  Iterative refinement has emerged as an effective paradigm for enhancing the capabilities of large language models (LLMs) on complex tasks. We propose Context-Wise Order-Agnostic Language Modeling (COrAL) to overcome these challenges. Our approach models multiple token dependencies within manageable context windows, enabling the model to perform iterative refinement internally.
  arXiv  Detail & Related papers  (2024-10-12T23:56:19Z)
• Natural Is The Best: Model-Agnostic Code Simplification for Pre-trained Large Language Models [6.646510073473929]
  We propose SlimCode, a model-agnostic code simplification solution for Large Language Models. SlimCode can improve the state-of-the-art technique by 9.46% and 5.15% in terms of MRR and BLEU score on code search and summarization.
  arXiv  Detail & Related papers  (2024-05-18T06:15:52Z)
• Reverse That Number! Decoding Order Matters in Arithmetic Learning [49.5504492920404]
  Our work introduces a novel strategy that reevaluates the digit order by prioritizing output from the least significant digit. Compared to the previous state-of-the-art (SOTA) method, our findings reveal an overall improvement of in accuracy while requiring only a third of the tokens typically used during training.
  arXiv  Detail & Related papers  (2024-03-09T09:04:53Z)
• Automated Sizing and Training of Efficient Deep Autoencoders using Second Order Algorithms [0.46040036610482665]
  We propose a multi-step training method for generalized linear classifiers. validation error is minimized by pruning of unnecessary inputs. desired outputs are improved via a method similar to the Ho-Kashyap rule.
  arXiv  Detail & Related papers  (2023-08-11T16:48:31Z)
• Quick Adaptive Ternary Segmentation: An Efficient Decoding Procedure For Hidden Markov Models [70.26374282390401]
  Decoding the original signal (i.e., hidden chain) from the noisy observations is one of the main goals in nearly all HMM based data analyses. We present Quick Adaptive Ternary (QATS), a divide-and-conquer procedure which decodes the hidden sequence in polylogarithmic computational complexity.
  arXiv  Detail & Related papers  (2023-05-29T19:37:48Z)
• CloudAttention: Efficient Multi-Scale Attention Scheme For 3D Point Cloud Learning [81.85951026033787]
  We set transformers in this work and incorporate them into a hierarchical framework for shape classification and part and scene segmentation. We also compute efficient and dynamic global cross attentions by leveraging sampling and grouping at each iteration. The proposed hierarchical model achieves state-of-the-art shape classification in mean accuracy and yields results on par with the previous segmentation methods.
  arXiv  Detail & Related papers  (2022-07-31T21:39:15Z)
• Neural Execution Engines: Learning to Execute Subroutines [29.036699193820215]
  We study the generalization issues at the level of numerical subroutines that comprise common algorithms like sorting, shortest paths, and minimum spanning trees. To generalize to unseen data, we show that encoding numbers with a binary representation leads to embeddings with rich structure once trained on downstream tasks like addition or multiplication.
  arXiv  Detail & Related papers  (2020-06-15T01:51:37Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.