</> Code with Papers by Syntology

Methods › Natural Language Processing › Autoencoding Transformers › Electric

Electric

167 papers tagged archive 2025-07-28

Introduced by Kevin Clark et al. in Pre-Training Transformers as Energy-Based Cloze Models

archive 2025-07-28 Description, source and code snippet are the archive's method entry.

Electric is an energy-based cloze model for representation learning over text. Like BERT, it is a conditional generative model of tokens given their contexts. However, Electric does not use masking or output a full distribution over tokens that could occur in a context. Instead, it assigns a scalar energy score to each input token indicating how likely it is given its context.

Specifically, like BERT, Electric also models $p_{\text {data }}\left(x_{t} \mid \mathbf{x}_{\backslash t}\right)$, but does not use masking or a softmax layer. Electric first maps the unmasked input 𝐱=[x₁, …, xₙ] into contextualized vector representations 𝐡(𝐱)=[𝐡₁, …, 𝐡ₙ] using a transformer network. The model assigns a given position t an energy score

E(𝐱)ₜ=𝐰ᵀ 𝐡(𝐱)ₜ

using a learned weight vector w. The energy function defines a distribution over the possible tokens at position t as

$$ p_{\theta}\left(x_{t} \mid \mathbf{x}_{\backslash t}\right)=\exp \left(-E(\mathbf{x})_{t}\right) / Z\left(\mathbf{x}_{\backslash t}\right) $$

=(exp(-E(𝐱)ₜ))/(∑_(x^' ∈𝒱) exp(-E(REPLACE(𝐱, t, x^'))ₜ))

where REPLACE(𝐱, t, x^') denotes replacing the token at position t with x^' and 𝒱 is the vocabulary, in practice usually word pieces. Unlike with BERT, which produces the probabilities for all possible tokens x^' using a softmax layer, a candidate x^' is passed in as input to the transformer. As a result, computing p_θ is prohibitively expensive because the partition function $Z_{\theta}\left(\mathbf{x}_{\backslash t}\right)$ requires running the transformer |𝒱| times; unlike most EBMs, the intractability of Z_θ(𝐱 \t) is more due to the expensive scoring function rather than having a large sample space.

PaperSource

Papers archive 2025-07-28

30 shown of 167, newest first. Repository counts are the archive's code-links table. A Syntology line states what Syntology ran from that paper's harvested code; it is per sample and not a correctness claim.

Tasks archive 2025-07-28

20 shown of 83 tasks the archive attaches to papers tagged with this method, by distinct papers. A task without a page in the catalog is plain text.

TaskPapers
Management27
Scheduling13
energy management13
Deep Reinforcement Learning11
Reinforcement Learning (RL)9
Time Series7
reinforcement-learning6
Anomaly Detection5
Model Predictive Control4
Transfer Learning4
ARC3
Autonomous Driving3
Combinatorial Optimization3
Computational Efficiency3
Decision Making3
Intrusion Detection3
Load Forecasting3
Time Series Analysis3
energy trading3
regression3

Usage over time archive 2025-07-28

Papers per year tagged with Electric: 2020 to 2025, peak 43 43 0 2020: 1 paper 2020 2021: 25 papers 2021 2022: 32 papers 2022 2023: 42 papers 2023 2024: 43 papers 2024 2025: 24 papers 2025
Papers per year the archive tags with this method, by the paper's archive date (167 dated). Bars are counts, not a trend claim.

Components: the archive holds no method-to-method composition, so PwC's Components table cannot be rebuilt; the Papers list carries no Results column for the same reason (the archive does not join its leaderboard rows to method tags).

Categories archive 2025-07-28

Autoencoding TransformersTransformersLanguage Models

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections