Pythia releases 16 identically trained LLMs with full checkpoints and data tools to study training dynamics, scaling, memorization, and bias in language models.
hub
C row S -Pairs: A Challenge Dataset for Measuring Social Biases in Masked Language Models
16 Pith papers cite this work. Polarity classification is still indexing.
16
Pith papers citing it
hub tools
citation-role summary
background 1
citation-polarity summary
roles
background 1polarities
background 1representative citing papers
Polar is a new cross-context benchmark showing LLM political bias measurements are not fixed but vary with country, issue, model, and language.
A Dutch BERT model encodes gender linearly by epoch 20 but does not dynamically update its representations when explicit female cues contradict learned stereotypical associations in short sentence templates.
The paper characterizes deductive stereotyping in LLMs and introduces Fair-GCG to discover injection phrases that improve fairness across benchmarks, reasoning, and real-world tasks.
H-SAL erases latent concepts from text profiles using self-descriptions as implicit debiasing signals and shows competitive performance on a new multi-domain Stack Exchange helpfulness benchmark.
LLMs correct only 34.8% of zero-shot annotation errors via prompting, and Definition-Specific Familiarity correlates positively with performance (partial r = +0.41) while memorization metrics do not.
LLMs show omissive bias by underrepresenting religious frameworks in responses to non-religious ethical questions relative to human expectations.
Counterfactual prompting effects on LLMs are often indistinguishable from those caused by meaning-preserving paraphrases, causing most previously reported demographic sensitivities to disappear under proper statistical comparison.
A methodological framework detects subtle group-associated linguistic biases in LLM outputs by generating controlled synthetic minimal pairs, abstracting n-grams, and ranking high-signal fragments with a PMI variant for expert review.
GMRL-BD detects untrustworthy topic boundaries for black-box LLMs by combining bias-diffusion on a Wikipedia KG with multi-agent RL, supported by a released dataset labeling biases in models like Llama2 and Qwen2.
The paper proposes the BADx metric to quantify persona-induced amplification of implicit intersectional biases in five LLMs, showing that context modulates bias beyond what static embedding tests capture.
UltraChat supplies 1.5 million high-quality multi-turn dialogues that, when used to fine-tune LLaMA, produce UltraLLaMA, which outperforms prior open-source chat models including Vicuna.
BLOOM is a 176B-parameter open-access multilingual language model trained on the ROOTS corpus that achieves competitive performance on benchmarks, with improved results after multitask prompted finetuning.
Galactica, a science-specialized LLM, reports higher scores than GPT-3, Chinchilla, and PaLM on LaTeX knowledge, mathematical reasoning, and medical QA benchmarks while outperforming general models on BIG-bench.
LLMs show minimal sociodemographic disparities in advice because they infer user demographics poorly from history; conversation topics are the main predictor and act as proxies for groups.
A survey reviewing benchmark data contamination in LLMs, its impact on evaluation, and alternative assessment approaches.
citing papers explorer
No citing papers match the current filters.