mlsanityData to 7 Oct 2026

Fantastic Semantics and Where to Find Them: Investigating Which Layers of Generative LLMs Reflect Lexical Semantics

arXiv 2403.01509 ↗First posted 3 Mar 2024cs.CLcited by 43

Not on the map. Why?

Abstract (arXiv)

Large language models have achieved remarkable success in general language understanding tasks. However, as a family of generative methods with the objective of next token prediction, the semantic evolution with the depth of these models are not fully explored, unlike their predecessors, such as BERT-like architectures. In this paper, we specifically investigate the bottom-up evolution of lexical semantics for a popular LLM, namely Llama2, by probing its hidden states at the end of each layer using a contextualized word identification task. Our experiments show that the representations in lower layers encode lexical semantics, while the higher layers, with weaker semantic induction, are responsible for prediction. This is in contrast to models with discriminative objectives, such as mask language modeling, where the higher layers obtain better lexical semantics. The conclusion is further supported by the monotonic increase in performance via the hidden states for the last meaningless symbols, such as punctuation, in the prompting strategy. Our codes are available at https://github.com/RyanLiut/LLM_LexSem.

Key facts. Cited by 43 papers in the map (data through 7 Oct 2026). Builds on 21 of its 31 references that are in the map. First posted on arXiv 3 Mar 2024.

Builds on

21 of 31 references.

References · 21

  1. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  2. 2020Language Models are Few-Shot Learnerscited by 32,199
  3. 2023Llama 2: Open Foundation and Fine-Tuned Chat Modelscited by 12,865
  4. 2013Distributed Representations of Words and Phrases and their Compositionalitycited by 3,462
  5. 2018Deep contextualized word representationscited by 2,285
  6. 2019Language Models as Knowledge Bases?cited by 1,922
  7. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,239
  8. 2022Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 smallcited by 1,130
  9. 2022In-context Learning and Induction Headscited by 930
  10. 2019How Contextual are Contextualized Word Representations? Comparing the Geometry of BERT, ELMo, and GPT-2 Embeddingscited by 770
  11. 2023A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPTcited by 612
  12. 2021Causal Abstractions of Neural Networkscited by 401
9 more
  1. 2018WiC: the Word-in-Context Dataset for Evaluating Context-Sensitive Meaning Representationscited by 370
  2. 2019What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language modelscited by 317
  3. 2023Label Words are Anchors: An Information Flow Perspective for Understanding In-Context Learningcited by 235
  4. 2017Advances in Pre-Training Distributed Word Representationscited by 207
  5. 2023ChatGPT: Jack of all trades, master of nonecited by 204
  6. 2019The Bottom-up Evolution of Representations in the Transformer: A Study with Machine Translation and Language Modeling Objectivescited by 139
  7. 2023Representation Of Lexical Stylistic Features In Language Models' Embedding Spacecited by 8
  8. 2019LIAAD at SemDeep-5 Challenge: Word-in-Context (WiC)cited by 7
  9. 2021Scalar Adjective Identification and Multilingual Rankingcited by 6

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2024: 3; 2025: 22; 2026: 18.

43 papers.

Sort:

By area:

  1. 2024The Remarkable Robustness of LLMs: Stages of Inference?cited by 138
  2. 2025SEAL: Steerable Reasoning Calibration of Large Language Models for Freecited by 65
  3. 2025GMSA: Enhancing Context Compression via Group Merging and Layer Semantic Alignmentcited by 20
  4. 2024Token Prepending: A Training-Free Approach for Eliciting Better Sentence Embeddings from LLMscited by 18
  5. 2025Language Model Uncertainty Quantification with Attention Chaincited by 13
  6. 2025Contrastive Prompting Enhances Sentence Embeddings in LLMs through Inference-Time Steeringcited by 13
  7. 2025A Comprehensive Study of Decoder-Only LLMs for Text-to-Image Generationcited by 10
  8. 2025Layerwise Recall and the Geometry of Interwoven Knowledge in LLMscited by 5
  9. 2025Spectral Insights into Data-Oblivious Critical Layers in Large Language Modelscited by 4
  10. 2026Tracing Moral Foundations in Large Language Modelscited by 4
  11. 2026Semantic Routing: Exploring Multi-Layer LLM Feature Weighting for Diffusion Transformerscited by 4
  12. 2025On-the-Fly Adaptive Distillation of Transformer to Dual-State Linear Attentioncited by 3
31 more
  1. 2025LayerCake: Token-Aware Contrastive Decoding within Large Language Model Layerscited by 3
  2. 2025TRIM: Token-wise Attention-Derived Saliency for Data-Efficient Instruction Tuningcited by 3
  3. 2025Hierarchical Token Prepending: Enhancing Information Flow in Decoder-based LLM Embeddingscited by 3
  4. 2026NITP: Next Implicit Token Prediction for LLM Pre-trainingcited by 3
  5. 2025From the New World of Word Embeddings: A Comparative Study of Small-World Lexico-Semantic Networks in LLMscited by 2
  6. 2025qa-FLoRA: Data-free query-adaptive Fusion of LoRAs for LLMscited by 2
  7. 2026Where meaning lives: Layer-wise accessibility of psycholinguistic features in encoder and decoder language modelscited by 2
  8. 2026From Associations to Activations: Comparing Behavioral and Hidden-State Semantic Geometry in LLMscited by 2
  9. 2025Emergent effects of scaling on the functional hierarchies within large language modelscited by 1
  10. 2025Beyond Prompting: An Efficient Embedding Framework for Open-Domain Question Answeringcited by 1
  11. 2025Self-ReS: Self-Reflection in Large Vision-Language Models for Long Video Understandingcited by 1
  12. 2025Plug-in and Fine-tuning: Bridging the Gap between Small Language Models and Large Language Modelscited by 1
  13. 2025Binary Autoencoder for Mechanistic Interpretability of Large Language Modelscited by 1
  14. 2025Learning to Look at the Other Side: A Semantic Probing Study of Word Embeddings in LLMs with Enabled Bidirectional Attentioncited by 1
  15. 2026Explainable AI: Context-Aware Layer-Wise Integrated Gradients for Explaining Transformer Modelscited by 1
  16. 2026Capsule Lens: Locating and Tracking Concept Geometry in Model Representationscited by 1
  17. 2024JuniperLiu at CoMeDi Shared Task: Models as Annotators in Lexical Semantics Disagreementscited by 0
  18. 2025Logo-LLM: Local and Global Modeling with Large Language Models for Time Series Forecastingcited by 0
  19. 2025R2-CoD: Understanding Text-Graph Complementarity in Relational Reasoning via Knowledge Co-Distillationcited by 0
  20. 2025HiCoLoRA: Addressing Context-Prompt Misalignment via Hierarchical Collaborative LoRA for Zero-Shot DSTcited by 0
  21. 2026PREGEN: Uncovering Latent Thoughts in Composed Video Retrievalcited by 0
  22. 2026Hiding in Plain Text: Detecting Concealed Jailbreaks via Activation Disentanglementcited by 0
  23. 2026Amnesia: Adversarial Semantic Layer Specific Activation Steering in Large Language Modelscited by 0
  24. 2026When Emotion Becomes Trigger: Emotion-style dynamic Backdoor Attack Parasitising Large Language Modelscited by 0
  25. 2026Model Unlearning Objectives Vary for Distinct Language Functionscited by 0
  26. 2026Cordyceps: Covert Control Attacks on LLMs via Data Poisoningcited by 0
  27. 2026ReverseEOL: Improving Training-free Text Embeddings via Text Reversal in Decoder-only LLMscited by 0
  28. 2026HydraQE: OSU's Submission for the IWSLT 2026 Speech Translation Metrics Shared Taskcited by 0
  29. 2026Technical Manual for a Toolkit for Measuring Contextual Individuation in Transformer Language Modelscited by 0
  30. 2026Selecting What Matters: Semantic Compression-Guided Selective Pooling for Long-Context Embeddingscited by 0
  31. 2026Uncovering Uncontrolled Repetition through Residual Stream Dynamicscited by 0