Surprisal minimization over goal-directed alternatives generated by language models provides the strongest account of production choices in open-ended dialogue compared to uniform information density or length-based costs.
hub
Cognition , year =
13 Pith papers cite this work. Polarity classification is still indexing.
13
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
support 1representative citing papers
HELM establishes a multi-metric evaluation covering 30 language models on 42 scenarios (16 core) to raise average scenario coverage from 17.9% to 96% under uniform conditions while releasing all prompts, completions, and a toolkit.
Transformers on impossible-language variants show gradual grammatical sensitivity loss but sharp long-sentence generation failures, supporting generative deficiency as a link to non-attestation.
Mamba's per-word timesteps significantly predict human reading times beyond GPT-2 surprisal in a naturalistic dataset.
Syntactic belief update via generalized Rényi divergence on syntactic trees predicts garden path reading times better than lexical surprisal.
Presents a minimal-pair dataset and reports that probing experiments show language models differentiate light-verb from full-verb uses even in minimal contexts with separable patterns by object type.
Sparse autoencoders applied to GPT-2 and Llama models recover semantic features accounting for 94% of peak brain encoding performance and map onto distinct cortical semantic regions across three languages.
Contextual curvature of LLM representational trajectories correlates with and causally modulates next-token entropy.
Later LLM layers align better with human cognitive effort in syntactic ambiguity than early layers do, indicating dual processing modes and complementary benefits from multi-layer probability updates.
Dutch LLMs display coherence illusions tracked by surprisal, with attention entropy identifying affected heads and a new energy metric quantifying discourse coherence.
Language models show idiom decomposability correlates weakly with human judgments, negatively with syntactic flexibility, and contributes most strongly to representation stabilization during training alongside surprisal and frequency.
Proposes CAC prompting to benchmark language models on syntactic and discourse properties of determiners against child acquisition data, finding large models approach but do not match human performance on both.
LLM surprisal and attention entropy replicate syncretism modulation of agreement attraction in English and German, align with null results in Turkish, and partially match Russian patterns.
citing papers explorer
-
Surprisal Minimisation over Goal-directed Alternatives Predicts Production Choice in Dialogue
Surprisal minimization over goal-directed alternatives generated by language models provides the strongest account of production choices in open-ended dialogue compared to uniform information density or length-based costs.
-
When transformers learn "impossible" languages, what do they learn?
Transformers on impossible-language variants show gradual grammatical sensitivity loss but sharp long-sentence generation failures, supporting generative deficiency as a link to non-attestation.
-
Timesteps of Mamba Align with Human Reading Times
Mamba's per-word timesteps significantly predict human reading times beyond GPT-2 surprisal in a naturalistic dataset.
-
Syntactic Belief Update as the Driver of Garden Path Processing Difficulty
Syntactic belief update via generalized Rényi divergence on syntactic trees predicts garden path reading times better than lexical surprisal.
-
Light or Full Verb? A Minimal-Pair Dataset for Probing Phraseological Competence in Language Models
Presents a minimal-pair dataset and reports that probing experiments show language models differentiate light-verb from full-verb uses even in minimal contexts with separable patterns by object type.
-
Sparse Autoencoders Map Brain-LLM Alignment onto Cortical Semantic Topography
Sparse autoencoders applied to GPT-2 and Llama models recover semantic features accounting for 94% of peak brain encoding performance and map onto distinct cortical semantic regions across three languages.
-
Representational Curvature Modulates Behavioral Uncertainty in Large Language Models
Contextual curvature of LLM representational trajectories correlates with and causally modulates next-token entropy.
-
Dual Alignment Between Language Model Layers and Human Sentence Processing
Later LLM layers align better with human cognitive effort in syntactic ambiguity than early layers do, indicating dual processing modes and complementary benefits from multi-layer probability updates.
-
When Context Misleads: Surprisal, Energy and Attention Entropy as Metrics of Coherence Illusions in LLMs
Dutch LLMs display coherence illusions tracked by surprisal, with attention entropy identifying affected heads and a new energy metric quantifying discourse coherence.
-
Rethinking the Idiomaticity Decomposability Hypothesis: Evidence from Distributional Learning
Language models show idiom decomposability correlates weakly with human judgments, negatively with syntactic flexibility, and contributes most strongly to representation stabilization during training alongside surprisal and frequency.
-
Measuring Form and Function in Language Models
Proposes CAC prompting to benchmark language models on syntactic and discourse properties of determiners against child acquisition data, finding large models approach but do not match human performance on both.
-
Quantifying the cross-linguistic effects of syncretism on agreement attraction
LLM surprisal and attention entropy replicate syncretism modulation of agreement attraction in English and German, align with null results in Turkish, and partially match Russian patterns.