When Only Time Will Tell: Interpreting How Transformers Process Local Ambiguities Through the Lens of Restart-Incrementality (2024.acl-long)
Copied to clipboard
| Challenge: | In incremental models, one interpretation is possible, but models that can revise can do so if the ambiguity is resolved. |
| Approach: | They propose an interpretable way to analyse incremental states in a bidirectional way . they propose to use a model that can update internal states to reflect the garden path effect . |
| Outcome: | The proposed model shows that it can perform revisions and recover if the label is incorrect. |
Similar Papers
Towards Incremental Transformers: An Empirical Analysis of Transformer Models for Incremental NLU (2021.emnlp-main)
Copied to clipboard
| Challenge: | Recent work attempts to apply incremental processing to NLUs but this is computationally expensive and does not scale efficiently for long sequences. |
| Approach: | They propose to apply Transformers incrementally via restart-incrementality by repeatedly feeding, to an unchanged model, increasingly longer input prefixes to produce partial outputs. |
| Outcome: | The proposed model has better incremental performance and faster inference speed compared to the standard Transformer and LT with restart-incrementality, at the cost of part of the non-incremental quality. |
Incremental Sentence Processing Mechanisms in Autoregressive Transformer Language Models (2025.naacl-long)
Copied to clipboard
| Challenge: | a recent study has found LMs focus on processing syntactic relations, but not structural information. |
| Approach: | They examine the mechanisms underlying garden path sentence processing in LMs . they use sparse autoencoders to identify interpretable features that determine which continuation . |
| Outcome: | The proposed model lacks syntactic features and shallow heuristics to perform incremental sentence processing. |
Incremental Processing in the Age of Non-Incremental Encoders: An Empirical Assessment of Bidirectional Models for Incremental NLU (2020.emnlp-main)
Copied to clipboard
| Challenge: | a number of languages are processed incrementally, but the best ones do not . we test five models on various datasets and compare their performance using three incremental evaluation metrics. |
| Approach: | They investigate how bidirectional LSTMs and Transformers behave under incremental interfaces . they propose to use bidirectional encoders in incremental mode while retaining non-incremental quality . |
| Outcome: | The proposed models perform better under incremental interfaces than the "omni-directional" BERT model, which achieves better non-incremental performance, but is impacted more by the incremental access. |
Beyond Semantics: How Temporal Biases Shapes Retrieval in Transformer and State-Space Models (2026.eacl-long)
Copied to clipboard
| Challenge: | In-context learning depends on what appears in the prompt and on when it appears. |
| Approach: | They construct prompts with repeated anchor tokens and average their predictions over hundreds of random permutations of the intervening context. |
| Outcome: | The proposed model predicts the same tokens over hundreds of permutations over time. |
Tug-of-war between idioms’ figurative and literal interpretations in LLMs (2026.eacl-long)
Copied to clipboard
| Challenge: | We find that idioms have non-compositional figurative interpretations that diverge from the idiomatic literal interpretation. |
| Approach: | They employ causal tracing to analyze how pretrained causal transformers deal with idiom ambiguity. |
| Outcome: | The proposed model leverages the idiom's context and refines it if it conflicts with the retrieved interpretation. |
Garden Path Recovery in Causal and Masked Language Models (2026.acl-srw)
Copied to clipboard
| Challenge: | a linguistics study of garden-path sentences shows that recovery dynamics are important for linguistic evaluation . causal models show larger within-model disambiguation effects than masked models overall . |
| Approach: | They propose to compare garden-path recovery in causal and masked language models . they use 100 English garden- path/control pairs spanning three constructions . |
| Outcome: | The proposed model shows that decoder-only models exhibit sharper disruption at the point of syntactic revision, while encoders appear comparatively buffered at the disambiguator due to right-context access. |
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. |
What Context Features Can Transformer Language Models Use? (2021.acl-long)
Copied to clipboard
| Challenge: | Recent studies show that transformer-based language models benefit from conditioning on contexts of hundreds to thousands of previous tokens. |
| Approach: | They propose to use lexical and structural information to ablate usable information in transformer language models. |
| Outcome: | The proposed model improves when conditioning on contexts of thousands of previous tokens. |
Bridging Internal Consistency and External Alignment: A Causal and Dynamic Interpretability Framework for LLM Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing interpretability methods focus on internal and external aspects of the model . existing explanations often focus on surface correlations or static dependencies . |
| Approach: | They propose a causal and dynamic interpretability framework for Large Language Models . they characterize backdoor-adjusted causal effects of generated prefix and prompt . |
| Outcome: | The proposed framework provides a unified causal view of internal consistency and external alignment in LLM generation dynamics. |
A Meta-Learning Perspective on Transformers for Causal Language Modeling (2024.findings-acl)
Copied to clipboard
| Challenge: | Mechanisms of the Transformer architecture for causal language modeling are not well understood. |
| Approach: | They propose a meta-learning view of the Transformer architecture when trained for a causal language modeling task by explicating an inner optimization process that may happen within the Transformer. |
| Outcome: | The proposed model is based on a self-attention mechanism and has been widely used in natural language processing, computer vision, and scientific discovery. |