Entity · paper

Instruction-Tuned Models Locally Reuse Human Syntax More Than Humans Do

paperactiveprovisionalinstruction-tuned-models-locally-reuse-human-syntax-more-than-humans-do-741b09f3·1 events·first seen 44h ago

Aliases: Instruction-Tuned Models Locally Reuse Human Syntax More Than Humans Do

Co-occurring entities

Gemma Llama

More like this (12)

Decomposing Factual Sycophancy in Language Models: How Size and Instruction Tuning Shape Robustness Prompt Design at Scale: How Format, Instruction Count, and Context Length Shape Instruction Adherence and Hallucination in Large Language Models Language Model Finetuning Construction-Driven Injection: Linguistically-Grounded Edit-Based Code-Mixing Fingerprints for Large Language Models Leveraging Instruction Tuning and Merging for Reasoning Model Adaptation Multimodal Continual Instruction Tuning On the Limits of Prompt-Conditioned Language Models as General-Purpose Learners Semantic Generative Tuning (SGT)Reasoning Language Models Fine-tuning GPT-2 from Human Preferences Measuring User's Mental Models of Speech Translation in Human-AI Collaboration Reproducing human biases in route choice using large language models: Toward scalable behavioral modeling

Recent events (1)

4arXiv · cs.CL·44h ago·source ↗

Instruction-tuned LLMs show stronger syntactic convergence toward human interlocutors than humans do

A new arXiv preprint measures syntactic convergence in 16 open-weight Llama and Gemma models (1B–70B, pretrained and instruction-tuned) by substituting model outputs into pre-existing human dialogues and measuring context-free grammar rule reuse. Every model showed greater CFG-rule overlap with the preceding human turn than a random baseline, and instruction-tuned models exceeded the syntactic convergence of the human responses they replaced. Instruction tuning increased local syntactic and semantic alignment with interlocutors, though it also raised baseline overlap with unrelated primes, suggesting a more globally accommodating output style rather than purely targeted priming.

Alignment and RLHF Gemma Instruction-Tuned Models Locally Reuse Human Syntax More Than Humans Do Llama