Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models

• URL: http://arxiv.org/abs/2504.15471v2
• Date: Fri, 25 Apr 2025 23:34:11 GMT
• Title: Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models
• Authors: Tyler A. Chang, Benjamin K. Bergen,
• Abstract summary: In Transformer language models, activation vectors transform from current token embeddings to next token predictions as they pass through the model.<n>To isolate a minimal form of this transformation, we identify language modelworks that make bigram predictions, naive next token predictions based only on the current token.<n>We find that bigramworks can be found in fully trained language models up to 1B parameters, and theseworks are critical for model performance even when they consist of less than 0.2% of model parameters.
• Score: 4.7936447642295406
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: In Transformer language models, activation vectors transform from current token embeddings to next token predictions as they pass through the model. To isolate a minimal form of this transformation, we identify language model subnetworks that make bigram predictions, naive next token predictions based only on the current token. We find that bigram subnetworks can be found in fully trained language models up to 1B parameters, and these subnetworks are critical for model performance even when they consist of less than 0.2% of model parameters. Bigram subnetworks are concentrated in the first Transformer MLP layer, and they overlap significantly with subnetworks trained to optimally prune a given model. Mechanistically, the bigram subnetworks often recreate a pattern from the full models where the first layer induces a sharp change that aligns activations with next token predictions rather than current token representations. Our results demonstrate that bigram subnetworks comprise a minimal subset of parameters that are both necessary and sufficient for basic next token predictions in language models, and they help drive the transformation from current to next token activations in the residual stream. These subnetworks can lay a foundation for studying more complex language model circuits by building up from a minimal circuit.

Related papers

• BriLLM: Brain-inspired Large Language Model [51.849486186292914]
  BriLLM is a non-Transformer, non-GPT, non-traditional machine learning input-output controlled generative language model.<n>We release the first BriLLM version in Chinese, with 4000 tokens, 32-dimensional node width, 16-token long sequence prediction ability, and language model prediction performance comparable to GPT-1.
  arXiv  Detail & Related papers  (2025-03-14T11:08:30Z)
• Scalable Language Models with Posterior Inference of Latent Thought Vectors [52.63299874322121]
  Latent-Thought Language Models (LTMs) incorporate explicit latent thought vectors that follow an explicit prior model in latent space.<n>LTMs possess additional scaling dimensions beyond traditional LLMs, yielding a structured design space.<n>LTMs significantly outperform conventional autoregressive models and discrete diffusion models in validation perplexity and zero-shot language modeling.
  arXiv  Detail & Related papers  (2025-02-03T17:50:34Z)
• Future Token Prediction -- Causal Language Modelling with Per-Token Semantic State Vector for Multi-Token Prediction [0.0]
  This research investigates a new pretraining method called Future Token Prediction (FTP) FTP generates embedding vectors for each token position that are linearly and expansively projected to a pseudo-sequence. On a toy, but complex, coding problem, FTP networks produce significantly better results than GPT networks.
  arXiv  Detail & Related papers  (2024-10-23T14:50:15Z)
• Semformer: Transformer Language Models with Semantic Planning [18.750863564495006]
  Next-token prediction serves as the dominant component in current neural language models. We introduce Semformer, a novel method of training a Transformer language model that explicitly models the semantic planning of response.
  arXiv  Detail & Related papers  (2024-09-17T12:54:34Z)
• On the Origins of Linear Representations in Large Language Models [51.88404605700344]
  We introduce a simple latent variable model to formalize the concept dynamics of the next token prediction. Experiments show that linear representations emerge when learning from data matching the latent variable model. We additionally confirm some predictions of the theory using the LLaMA-2 large language model.
  arXiv  Detail & Related papers  (2024-03-06T17:17:36Z)
• Unveiling Multilinguality in Transformer Models: Exploring Language Specificity in Feed-Forward Networks [12.7259425362286]
  We investigate how multilingual models might leverage key-value memories. For autoregressive models trained on two or more languages, do all neurons (across layers) respond equally to all languages? Our findings reveal that the layers closest to the network's input or output tend to exhibit more language-specific behaviour compared to the layers in the middle.
  arXiv  Detail & Related papers  (2023-10-24T06:45:00Z)
• Why do Nearest Neighbor Language Models Work? [93.71050438413121]
  Language models (LMs) compute the probability of a text by sequentially computing a representation of an already-seen context. Retrieval-augmented LMs have shown to improve over standard neural LMs, by accessing information retrieved from a large datastore.
  arXiv  Detail & Related papers  (2023-01-07T11:12:36Z)
• Pre-Training a Graph Recurrent Network for Language Representation [34.4554387894105]
  We consider a graph recurrent network for language model pre-training, which builds a graph structure for each sequence with local token-level communications. We find that our model can generate more diverse outputs with less contextualized feature redundancy than existing attention-based models.
  arXiv  Detail & Related papers  (2022-09-08T14:12:15Z)
• Pretraining Graph Neural Networks for few-shot Analog Circuit Modeling and Design [68.1682448368636]
  We present a supervised pretraining approach to learn circuit representations that can be adapted to new unseen topologies or unseen prediction tasks. To cope with the variable topological structure of different circuits we describe each circuit as a graph and use graph neural networks (GNNs) to learn node embeddings. We show that pretraining GNNs on prediction of output node voltages can encourage learning representations that can be adapted to new unseen topologies or prediction of new circuit level properties.
  arXiv  Detail & Related papers  (2022-03-29T21:18:47Z)
• Revisiting Simple Neural Probabilistic Language Models [27.957834093475686]
  This paper revisits the neural probabilistic language model (NPLM) ofcitetBengio2003ANP. When scaled up to modern hardware, this model performs much better than expected on word-level language model benchmarks. Inspired by this result, we modify the Transformer by replacing its first self-attention layer with the NPLM's local concatenation layer.
  arXiv  Detail & Related papers  (2021-04-08T02:18:47Z)
• Addressing Some Limitations of Transformers with Feedback Memory [51.94640029417114]
  Transformers have been successfully applied to sequential, auto-regressive tasks despite being feedforward networks. We propose the Feedback Transformer architecture that exposes all previous representations to all future representations. We demonstrate on a variety of benchmarks in language modeling, machine translation, and reinforcement learning that the increased representation capacity can create small, shallow models with much stronger performance than comparable Transformers.
  arXiv  Detail & Related papers  (2020-02-21T16:37:57Z)

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.