Semantic Softmax aggregates probabilities from semantic synonyms around target labels to correct renormalization bias in zero-shot LLM classification, lowering calibration error and raising AUROC and F1.
Proceedings of the 34th International Conference on Machine Learning , pages =
7 Pith papers cite this work. Polarity classification is still indexing.
7
Pith papers citing it
citation-role summary
method 1
citation-polarity summary
years
2026 7roles
method 1polarities
use method 1representative citing papers
DITTO uses RL with verbal feedback to train LLMs for human behavior simulation, reporting 36% average gains over base models and outperforming GPT-5.4 on 6 of 10 SOUL benchmark tasks.
Clarification-seeking in LLM agents amplifies prompt injection attack success from ~2% to over 30% across ten frontier models in a new 728-scenario benchmark.
Introduces a role-separated audit for release-side risk in conformal triage under prevalence shift and applies it to an NSCLC pilot showing that reduced review rates can release event-positive cases.
Mainstream UQ for LLMs reduces to unsupervised clustering of internal generation consistency and therefore cannot detect confident hallucinations or provide reliable safety signals.
Neural feature maps create expressive kernels that enable fast, scalable, and consistent exact Gaussian process inference for regression and classification.
Outcome-level RL with binary or composite rewards improves compositional generalization over supervised fine-tuning by avoiding overfitting to frequent training patterns.
citing papers explorer
-
The Silent Vote: Improving Zero-Shot LLM Reliability by Aggregating Semantic Neighborhoods
Semantic Softmax aggregates probabilities from semantic synonyms around target labels to correct renormalization bias in zero-shot LLM classification, lowering calibration error and raising AUROC and F1.
-
Reinforcing Human Behavior Simulation via Verbal Feedback
DITTO uses RL with verbal feedback to train LLMs for human behavior simulation, reporting 36% average gains over base models and outperforming GPT-5.4 on 6 of 10 SOUL benchmark tasks.
-
ASPI: Seeking Ambiguity Clarification Amplifies Prompt Injection Vulnerability in LLM Agents
Clarification-seeking in LLM agents amplifies prompt injection attack success from ~2% to over 30% across ten frontier models in a new 728-scenario benchmark.
-
A Deployment Audit of Release-Side Risk in Conformal Triage under Prevalence Shift
Introduces a role-separated audit for release-side risk in conformal triage under prevalence shift and applies it to an NSCLC pilot showing that reduced review rates can release event-positive cases.
-
Position: Uncertainty Quantification in LLMs is Just Unsupervised Clustering
Mainstream UQ for LLMs reduces to unsupervised clustering of internal generation consistency and therefore cannot detect confident hallucinations or provide reliable safety signals.
-
Scalable Gaussian process inference via neural feature maps
Neural feature maps create expressive kernels that enable fast, scalable, and consistent exact Gaussian process inference for regression and classification.
-
Reinforcement Learning for Compositional Generalization with Outcome-Level Optimization
Outcome-level RL with binary or composite rewards improves compositional generalization over supervised fine-tuning by avoiding overfitting to frequent training patterns.