LLMs maintain surface syntax for novel CFGs but fail to preserve semantics under recursion and branching, relying on keyword bootstrapping rather than pure symbolic reasoning.
arXiv preprint arXiv:2408.11061 (2024)
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
representative citing papers
Empirical benchmarks on four SE tasks show grammar-constrained decoding and TTMG eliminate most syntax errors in LLM outputs while structural and semantic errors persist and cascade in downstream tools.
ProtoMedAgent formalizes multimodal clinical reporting as iterative zero-gradient test-time optimization over a neuro-symbolic bottleneck with k-anonymity and ℓ-diversity privacy gate, reporting 91.2% faithfulness versus 46.2% for standard RAG on a 4,160-patient cohort.
Physicians use substantially more risk-focused framing in counseling notes for repeat cesarean than for VBAC among patients clinically eligible for both.
An end-to-end LLM framework refines natural language into valid PDDL domains and problems via hardcoded and dynamic agents, generates plans with standard engines, and returns readable output.
LLM system using LoRA, in-context learning, and ensembles achieves top performance on Chinese essay rhetoric recognition task.
citing papers explorer
-
Diagnosing CFG Interpretation in LLMs
LLMs maintain surface syntax for novel CFGs but fail to preserve semantics under recursion and branching, relying on keyword bootstrapping rather than pure symbolic reasoning.
-
Empirical Study for Structured Output Control in LLMs for Software Engineering
Empirical benchmarks on four SE tasks show grammar-constrained decoding and TTMG eliminate most syntax errors in LLM outputs while structural and semantic errors persist and cascade in downstream tools.
-
ProtoMedAgent: Multimodal Clinical Interpretability via Privacy-Aware Agentic Workflows
ProtoMedAgent formalizes multimodal clinical reporting as iterative zero-gradient test-time optimization over a neuro-symbolic bottleneck with k-anonymity and ℓ-diversity privacy gate, reporting 91.2% faithfulness versus 46.2% for standard RAG on a 4,160-patient cohort.
-
Implicit Framing in Obstetric Counseling Notes: A Grounded LLM Pipeline on a VBAC-Eligible Cohort
Physicians use substantially more risk-focused framing in counseling notes for repeat cesarean than for VBAC among patients clinically eligible for both.
-
End-to-end PDDL Planning with Hardcoded and Dynamic Agents
An end-to-end LLM framework refines natural language into valid PDDL domains and problems via hardcoded and dynamic agents, generates plans with standard engines, and returns readable output.
-
Chinese Essay Rhetoric Recognition Using LoRA, In-context Learning and Model Ensemble
LLM system using LoRA, in-context learning, and ensembles achieves top performance on Chinese essay rhetoric recognition task.
- Synthetic Homes: A Multimodal Generative AI Pipeline for Residential Building Data Generation under Data Scarcity