Reconsidering the Past: Optimizing Hidden States in Language Models (2021.findings-emnlp)

Copied to clipboard

Challenge: In this paper, we present Hidden-State Optimization (HSO) for language models at inference time.
Approach: They propose a method that uses the log-probability gradient to update hidden states rather than the model parameters to improve the performance of transformer language models.
Outcome: The proposed method improves performance of transformer language models at inference time.

Similar Papers

Hidden State Variability of Pretrained Language Models Can Guide Computation Reduction for Transfer Learning (2022.findings-emnlp)

Copied to clipboard

Challenge: Existing approaches to transfer a pretrained language model include fine-tuning all the parameters in the language model and adapting all its subsets.
Approach: They propose to select layers based on the variability of their hidden states given a task-specific corpus.
Outcome: The proposed model reduces the computational cost of transfer learning methods without sacrificing performance.
HyPe: Better Pre-trained Language Model Fine-tuning with Hidden Representation Perturbation (2023.acl-long)

Copied to clipboard

Challenge: Existing techniques to fine-tune pre-trained language models on downstream tasks are inadequate.
Approach: They propose a technique to perturb hidden Transformers representations by enhancing generalization of hidden representations from different layers.
Outcome: The proposed technique outperforms vanilla fine-tuning and enhances generalization of hidden representations from different layers.
Token-wise Decomposition of Autoregressive Language Model Hidden States for Analyzing Model Predictions (2023.acl-long)

Copied to clipboard

Challenge: Recent work on why Transformer-based large language models make predictions has made their behavior opaque due to the complexity of the computations performed within each layer.
Approach: They propose a linear decomposition of final hidden states from autoregressive language models based on each initial input token, which is exact for virtually all contemporary Transformer architectures.
Outcome: The proposed method analyzes the influence of input tokens on model probabilities over a sequence of upcoming words with only one forward pass from the model.
Improving the Language Understanding Capabilities of Large Language Models Using Reinforcement Learning (2025.findings-emnlp)

Copied to clipboard

Challenge: Instruction-fine-tuned large language models (LLMs) under 14B parameters underperform on NLU tasks . we explore a framework to improve the NLU capabilities of LLMs .
Approach: They propose to use Proximal Policy Optimization to improve NLU capabilities . they frame NLU as a reinforcement learning environment and optimize for reward signals .
Outcome: The proposed framework outperforms supervised fine-tuning on GLUE and superGLUE tasks.
Scaling Hidden Markov Language Models (2020.emnlp-main)

Copied to clipboard

Challenge: Hidden Markov models are a fundamental tool for sequence modeling that separates the hidden state from the emission structure.
Approach: They propose methods for scaling hidden Markov models to massive state spaces while maintaining efficient exact inference and effective regularization.
Outcome: The proposed methods are much more accurate than previous HMMs and n-gram-based methods, making progress towards the performance of state-of-the-art NN models.
Hidden States as Early Signals: Step-level Trace Evaluation and Pruning for Efficient Test-Time Scaling (2026.findings-acl)

Copied to clipboard

Challenge: Existing approaches to speed up parallel scaling have relied on similarity-based or confidence-based pruning, but these signals do not reliably indicate trace quality.
Approach: They propose a pruning framework that evaluates reasoning steps using hidden states and dynamically prunes unpromising traces during generation.
Outcome: The proposed framework reduces end-to-end inference latency by 45%–70% on average compared to self-consistency while improving reasoning accuracy.
An Empirical Study on Hyperparameter Optimization for Fine-Tuning Pre-trained Language Models (2021.acl-long)

Copied to clipboard

Challenge: In the recent years, pre-trained language models have achieved great success in the NLP community.
Approach: They propose two general strategies and an experimental procedure to troubleshoot HPO’s failure cases.
Outcome: The proposed methods outperform grid search on two state-of-the-art language models using the same time budget and overfitting.
Better Exploiting Latent Variables in Text Modeling (P19-1)

Copied to clipboard

Challenge: Consistent gains in performance on two datasets, Penn Treebank and Yahoo, indicate the generalizability of our method.
Approach: They propose a method to exploit latent variables through hidden state averaging by sampling latent variable multiple times at a gradient step.
Outcome: The proposed method shows consistent gains on two datasets showing that it is generalizable.
Multi-armed bandits for resource efficient, online optimization of language model pre-training: the use case of dynamic masking (2023.findings-acl)

Copied to clipboard

Challenge: Using a Bayesian optimization framework, we pre-train Transformer-based language models (TLMs) using a multi-armed bandit framework requires high computational resources and introduces many unresolved design choices.
Approach: They propose a Bayesian optimization framework for resource efficient pre-training of Transformer-based language models.
Outcome: The proposed framework achieves lower MLM loss in fewer epochs, across settings, while avoiding expensive hyperparameter grid search.
DenseSSM: State Space Models with Dense Hidden Connection for Efficient Large Language Models (2025.naacl-long)

Copied to clipboard

Challenge: Large language models (LLMs) face excessive computational and memory requirements due to the commonly used Transformer architecture.
Approach: They propose a method to enhance the flow of hidden information between layers in large language models by selectively integrating shallow-layer hidden states into deeper layers.
Outcome: The proposed method maintains parallelizability and inference efficiency of SSMs while significantly boosting performance on public benchmarks.

What is GenGO?

GenGO is an NLP powered publication search system. It currenctly indexes 30k+ papers from ACL Anthology, and implements multi-aspect summarization, semantic search, and more!

Information

About
Limitations