REVIEW 5 cited by
Language Models Use Trigonometry to Do Addition
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
abstract
Mathematical reasoning is an increasingly important indicator of large language model (LLM) capabilities, yet we lack understanding of how LLMs process even simple mathematical tasks. To address this, we reverse engineer how three mid-sized LLMs compute addition. We first discover that numbers are represented in these LLMs as a generalized helix, which is strongly causally implicated for the tasks of addition and subtraction, and is also causally relevant for integer division, multiplication, and modular arithmetic. We then propose that LLMs compute addition by manipulating this generalized helix using the "Clock" algorithm: to solve $a+b$, the helices for $a$ and $b$ are manipulated to produce the $a+b$ answer helix which is then read out to model logits. We model influential MLP outputs, attention head outputs, and even individual neuron preactivations with these helices and verify our understanding with causal interventions. By demonstrating that LLMs represent numbers on a helix and manipulate this helix to perform addition, we present the first representation-level explanation of an LLM's mathematical capability.
Forward citations
Cited by 5 Pith papers
-
Temporal Preference Concepts and their Functions in a Large Language Model
Temporal preference in Qwen3-4B-Instruct-2507 localizes to layers 17–35 (especially L24 attention), has curved residual-stream geometry, is behaviorally unstable, and can be bidirectionally steered.
-
Are Arithmetic Heuristic Neurons Form-Invariant? A Mechanistic Analysis of Symbols, Text, and Code in LLMs
Llama-3 arithmetic is computed by a shared neuron set across symbolic, text, and code formats, and cross-format failures are activation-state differences, not distinct circuits.
-
Singular Vectors of Attention Heads Align with Features
Under linear feature representations, attention-head singular vectors align with attended features under near-isotropy, and this predicts sparse attention decomposition seen in Pythia and GPT-2.
-
Modular Arithmetic: Language Models Solve Math Digit by Digit
LLMs perform 3-digit addition and subtraction via digit-position-specific MLP circuits that can be intervened upon to change individual output digits.
-
Prompting as Scientific Inquiry
Position paper arguing that prompting LLMs is a form of behavioral science and should be recognized as a core scientific method alongside mechanistic interpretability.
Discussion (0). Sign in to comment.