arXiv Computation and Language

Representational drift in a code model: behaviour settles, which neurons carry it does not

The study investigates representational drift in the OpenCoder‑1.5B language model by tracking how neurons encode 169 Python constructs during pretraining on two trillion tokens. While the loss for 16 grammar‑fixing constructs quickly reaches a floor, the specific neurons that carry these constructs continue to change throughout training, with no stable carrier set emerging even at the final checkpoint. The authors find that removing the identified carriers incurs a measurable performance penalty, yet the carrier sets remain dynamic, challenging the hypothesis that they settle over time.

arXiv Computation and Language
Nov 9

Calibrated concept profiles: making the reading of a model's neurons a measurement

The paper introduces calibrated concept profiles, a method that treats direct neuron readings as measurable quantities. By sampling prompts from a defined distribution and generating two independent batteries, the authors compute a per‑neuron profile expressed in the neuron’s own activation spread, complete with noise units and a minimum resolvable change. The approach is applied to OpenCoder‑1.5B across checkpoints spanning two trillion training tokens, providing unit‑scaled, chance‑level‑aware, and checkpoint‑specific construct attribution for each neuron.

By Piotr Wilam
arXiv Machine Learning
Oct 5

Neural Data Needs Semantic Tokenization: Behavioral Events as Boundaries of Session-Transferable Tokens

The paper introduces Tokenization with States (TWS), a method that segments neural recordings into behaviorally meaningful states defined by task events and represents each state as a token based on population geometry, bypassing neuron- or session-specific embeddings. TWS successfully decodes movement from regime boundaries that carry no target information and transfers across species and recording modalities with only a linear probe. The approach demonstrates that tokens derived from shared population dynamics can generalize across sessions, outperforming per-neuron foundation models that fail to generalize.

By Sangyoon Bae, Jiook Cha
arXiv Computation and Language
Sep 25

What a Cross-Model Fixed-Point Census Can and Cannot Arbitrate About Repetition

The paper investigates neural text degeneration by measuring the fixed‑point structure of short‑window argmax maps across 17 pretrained models, using 96 random two‑token starts without prompts. It finds a stable four‑way classification that varies across model families and scales, with some models funneling to a single endpoint token while others do not, and shows that this behavior is not solely determined by training data or corpus frequency. The study demonstrates that repetition phenomena are not uniformly explained by either training data or network architecture alone, highlighting the complexity of neural text generation dynamics.

By Nicol\'as Vera Z\'u\~niga
arXiv AI
Oct 16

Learning to Read Out: Unembedding Dynamics in Language Model Pretraining

The paper investigates when a language model actually learns task information versus when its readout (unembedding matrix) learns to express that information. Using a new parameter‑trajectory cross‑coding method, the authors find that hidden states can encode task data early in training, but the readout reorganizes only in a narrow window (0.4–1.1% of Pythia’s pretraining). By varying the readout learning rate, they show that faster readout learning reduces the gap between hidden‑state encoding and output expression without affecting final performance.

By Matteo He, William F. Shen, Alex Iacob, Andrej Jovanovic, Xinchi Qiu, Nicholas D. Lane
arXiv Machine Learning
6d ago

Are Causal Neuron Circuits Unique and Transferable?

The paper investigates whether groups of neurons that influence a model’s answer are uniquely necessary and transferable across inputs. By exhaustively testing all 4,096 subsets of 12 MLP neurons in Qwen3 and other models, the authors find that only a few neurons are often sufficient to alter predictions, that many subsets can independently flip answers, and that the effect is strongest on the same input chain but weaker on different chains. The study shows that causal neuron circuits are not uniquely defined and that their influence can vary with context and model architecture.

By \"Omer Tekin Yavuz (\.Ileti\c{s}im Yaz{\i}l{\i}m)
arXiv Machine Learning
Oct 23

Delta N-gram: Forward-Only Continual Writing into a Language Model's Native N-gram Memory

Delta N-gram introduces a forward‑only write path into a language model’s native hashed N‑gram table, preserving the pretrained table while adding evidence through a sparse patch Δ. The method uses a 4.1 M‑parameter writer that emits gated targets per event and applies a block‑averaged delta rule to keep each row a convex combination of its old value and the new targets. Experiments on a 0.6 B prototype and Qwen3.8‑Flash‑Next show that the delta writer maintains more document‑specific memory and reduces drift compared to online gradient writes, achieving stable performance over thousands of writes. whyItMatters":"The approach allows continual updates to a language model’s memory without altering its pretrained parameters, enabling efficient, stable, and document‑specific knowledge retention during inference."

By Hao Wu
arXiv AI
Oct 12

NeuronSpark: Selective Spiking State Space Models with Adaptive Computation for Language Modeling

NeuronSpark is a self‑attention‑free language model that uses input‑dependent spiking membrane dynamics, allowing each neuron’s decay rate, gain, and firing threshold to adapt to the current token. This selective spiking approach enables billion‑parameter pretraining from random initialization without relying on backbone transfer, conversion, or distillation, achieving competitive zero‑shot performance and a fast–slow hierarchy. On a 1.1B‑parameter, 2.1B‑token training run, NeuronSpark attains a 40.0 eight‑task zero‑shot average, surpasses comparable‑budget Pythia/Mamba checkpoints, and demonstrates significant energy and cycle reductions on FPGA implementations.

By Zhengzheng Tang