Rare Tokens Degenerate All Tokens: Improving Neural Text Generation via Adaptive Gradient Gating for Rare Token Embeddings (2022.acl-long)
Copied to clipboard
| Challenge: | Recent studies have determined that the learned token embeddings of large-scale neural language models are degenerated to be anisotropic with a narrow-cone shape. |
| Approach: | They propose a method to degenerate the learning gradient for rare token embeddings by gating the specific part of the gradient for all tokens during training stage. |
| Outcome: | The proposed method improves the performance of the models but lacks the training dynamics needed to solve the representation degeneration problem. |
Similar Papers
Revisiting Representation Degeneration Problem in Language Modeling (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Language modeling is a fundamental task in natural language processing, applications include machine translation, image captioning and speech recognition. |
| Approach: | They propose a cosine regularization method to solve the representation degeneration problem by analyzing the limitations of the proposed method and then propose an alternative regularization technique to tackle the problem. |
| Outcome: | The proposed method is effective in language modeling and image captioning. |
Representation Degeneration Problem in Prompt-based Models for Natural Language Understanding (2024.lrec-main)
Copied to clipboard
| Challenge: | Prompt-based fine-tuning (PF) models have shown improved performance on few-shot natural language understanding benchmarks. |
| Approach: | They propose a framework to alleviate anisotropy in the embedding space by aligning with pre-trained language models' training objective. |
| Outcome: | The proposed method outperforms mainstream methods on many NLU benchmarks. |
Empowering Diffusion Models on the Embedding Space for Text Generation (2024.naacl-long)
Copied to clipboard
| Challenge: | Recent work adapts diffusion models to textual data by diffusing on the embedding space. |
| Approach: | They propose an embedding diffusion model based on Transformer to solve the problem of embeddable space and denoising model. |
| Outcome: | The proposed model is more efficient than previous methods on seminal text generation tasks and is superior to existing models. |
Dynamic Meta-Embeddings for Improved Sentence Representations (D18-1)
Copied to clipboard
| Challenge: | A sprawling literature has emerged about what word embeddings are most useful for which tasks . word embed-ding is a technique that can be used to learn word-level meaning representations for a variety of tasks. |
| Approach: | They propose a method for supervised learning of embedding ensembles that leads to state-of-the-art performance on a variety of tasks. |
| Outcome: | The proposed method leads to state-of-the-art performance on a variety of tasks. |
Too Much in Common: Shifting of Embeddings in Transformer Language Models and its Implications (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing studies have shown that word embeddings do not occupy a narrow cone, but rather drift in common directions. |
| Approach: | They show that anisotropy can be restored using a simple transformation of word embeddings. |
| Outcome: | The proposed model can restore anisotropy using a simple transformation. |
A Cluster-based Approach for Improving Isotropy in Contextual Embedding Space (2021.acl-short)
Copied to clipboard
| Challenge: | Existing approaches to address the representation degeneration problem in contextual embedding spaces require a learning process to retrain models with additional objectives. |
| Approach: | They propose a local cluster-based method to address the representation degeneration problem in contextual embedding spaces by removing local dominant directions from verb representations. |
| Outcome: | The proposed method improves CWRs performance on semantic tasks by removing dominant directions of verb representations. |
Reward Gaming in Conditional Text Generation (2023.acl-long)
Copied to clipboard
| Challenge: | Recent work has used reward functions learned from human annotations to align conditional text generation models with desired behaviors. |
| Approach: | They propose to use reinforcement learning to train conditional text generation models with reward functions learned from human annotations to align outputs with desired behaviors. |
| Outcome: | The proposed framework improves the quality of generated summaries by using saliency and faithfulness metrics. |
Breaking the Generator Barrier: Disentangled Representation for Generalizable AI-Text Detection (2026.acl-long)
Copied to clipboard
| Challenge: | a rapid proliferation of large language models (LLMs) generate text that increasingly resembles human writing . this makes it difficult to capture subtle cues that distinguish AI-generated content from human-written content . |
| Approach: | They propose a framework that disentangles AI-detection semantics from generator-aware artifacts by latent encoding and perturbation-based regularization. |
| Outcome: | The proposed framework disentangles AI-detection semantics from generator-aware artifacts on 20 representative LLMs across 7 categories. |
Stable Language Model Pre-training by Reducing Embedding Variability (2024.emnlp-main)
Copied to clipboard
| Challenge: | Stable pre-training is essential for achieving better-performing language models, but tracking pre-train stability is impractical due to high computational costs. |
| Approach: | They propose to use Token Embedding Variability as a proxy to estimate pre-training stability. |
| Outcome: | The proposed method improves stability and lowers perplexities even at deeper layer counts. |
Exploring the Hidden Capacity of LLMs for One-Step Text Generation (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) can reconstruct surprisingly long texts via autoregressive generation from just one trained input embedding. |
| Approach: | They show that large language models can reconstruct surprisingly long texts via autoregressive generation from just one trained input embedding. |
| Outcome: | The proposed model can generate hundreds of accurate tokens in one token-parallel forward pass, when provided with only two learned embeddings. |