arXiv Computation and Language

The Communication Map of a Transformer

The paper introduces the Communication Map, a method that charts every potential communication channel in a transformer model using only its weights. It generalizes previous coupling metrics into a single coefficient covering all 18 connection classes, revealing that 70‑89% of head pairs are non‑randomly oriented and identifying strong or avoiding couplings. The authors demonstrate the map’s utility by recovering known induction circuits and uncovering a two‑dimensional stream subspace whose removal eliminates induction capabilities across several models.

arXiv Computation and Language
Oct 28

Mechanistic Failure Modes of Weight Pruning in Transformer Language Models

The study investigates how pruning affects internal mechanisms of GPT‑2 by systematically pruning attention heads and MLP neurons using magnitude and random importance criteria at 3%, 50%, and 70% sparsity. Using the indirect object identification (IOI) circuit as a ground truth, the authors measure circuit integrity with activation patching, path patching, and direct logit attribution, revealing two distinct failure modes: magnitude head pruning essentially zeros the circuit, while magnitude neuron pruning at 30% inverts it, causing the model to prefer incorrect names. The results show that conventional output metrics like perplexity can miss significant mechanistic damage, underscoring the need for circuit‑level causal analysis when evaluating compressed language models.

By Nabeelah Faumi, Deshan Sumanathilaka
arXiv Computation and Language
Sep 10

Through the Looking Glass: Directly Reading and Writing Transformers

The paper investigates how many transformer components influence a token prediction by measuring the absolute contribution of each unit and channel to the logit. It finds that thousands of components contribute to a single prediction, yet a small subset—often just dozens—carries the majority of the predictive mass. Across models ranging from 124 M to 7 B parameters, the proportion of the model involved in a prediction remains around one to three percent, independent of size, and the study demonstrates that specific components can be directly read and written to modify model behavior without additional training.

By Mark Oskin
arXiv AI
Sep 25

Every Component Is a Lookup: One Linear Graph for Interaction, Composition and Attribution

The paper proposes that two architectural assumptions—(1) attention and MLPs share a key‑value form <phi(S)>U, and (2) components read from an additive residual stream—are sufficient to answer three interpretability questions: component interaction, information routing, and token attribution. By treating these selections as a computational graph, the authors develop Unpack, a backward attribution method that validates interaction scores, recovered routes, and token attribution against established tests across models ranging from 160M to 6.9B parameters. The study also shows that contribution and causal effect can differ, with a recognizable signature in how components change when a task is removed.

By Po-Kai Chen, Aske Plaat, Niki van Stein
arXiv AI
Aug 20

First-Token Broadcasters: Mechanistic Origins of Language Identity and Distributed Robustness in Transformers

The paper introduces Language Identity Head Ablation (LIHA), a causal method that zeroes individual attention heads in transformer models to measure language switch rates across multilingual prompts. Applying LIHA to GPT‑2 reveals a small set of first‑token broadcaster heads—most notably L6H1—that persistently attend to the initial prompt token and propagate language signals throughout generation, with compensatory head recruitment occurring hierarchically in higher layers. A controlled comparison between Qwen2.5‑1.5B‑Base and Qwen2.5‑1.5B‑Instruct shows that instruction tuning concentrates language‑identity influence in early layers, while experiments with Chinese and Russian confirm script‑specific first‑token broadcasting at layer 0.

By Arjun Pillai, Christian Hoang, Anjelo Jann Laroza
arXiv AI
Jun 2

When Do Attention Circuits Form? Developmental Trajectories of Capability and Attention-Sink Emergence Across Three 1B-ClassArchitectures

arXiv:2606. 02378v1 Announce Type: cross Abstract: We track the developmental trajectory of attention-head circuit formation across three 1B-class language models spanning two architecture families (dense transformer, mixture-of-experts) and two pretraining corpora (The Pile, DCLM): Pythia 1B, OLMo 1B-0724-hf, and OLMoE 1B-7B-0924.

By Yongzhong Xu
arXiv Machine Learning
Jun 5

Pattern Selectivity is Not Task-Causal Structure: A Cross-Architecture Mechanistic Study of Composed-Task Circuits in 1B-Class Language Models

arXiv:2606. 05378v1 Announce Type: new Abstract: We test whether a single screen-and-ablate recipe -- identify attention-head circuits by task-pattern selectivity, then verify by causal ablation against a matched-random null -- produces consistent mechanistic claims across model families.

By Yongzhong Xu
arXiv Computation and Language
Sep 4

How Much Do Circuits Tell Us? Measuring the Consistency and Specificity of Language Model Circuits

The study investigates the consistency and specificity of language model circuits across six tasks and five models, focusing on component-level (attention heads and MLP blocks) and neuron-level circuits. Component-level circuits are highly consistent and causally important but lack task specificity, as ablating a circuit for one task similarly harms performance on other tasks. Neuron-level circuits show higher task specificity but lower consistency, with overlap mainly between closely related tasks. The analysis of Llama‑3.2‑3B reveals that shared components are predominantly MLP blocks, while attention heads act as generic attention‑sink heads.

By Michael Li, Nishant Subramani
arXiv AI
Oct 27

Relational Composition as Parallel Transport: What Per-Entity Frames Force in Attention

The paper investigates how treating each entity in a graph or language model as having its own coordinate frame—viewed as a symmetry—constrains the form of attention mechanisms. It shows that with orthogonal transports, invariant attention scores depend only on norms and transported inner products, while equivariant messages lie in the span of the entity vectors and their transported counterparts. The authors demonstrate that standard attention corresponds to a flat shared frame, and that incorporating per‑entity frames allows relations to be represented as operators that compose via parallel transport, yielding empirical gains on synthetic relational tasks and natural language benchmarks.

By Aakash Sarkar, Ananya Mukherjee, Motoshige Sato