Papers with systematic study
Theory-Grounded Measurement of U.S. Social Stereotypes in English Language Models (2022.naacl-main)
Copied to clipboard
| Challenge: | Pre-trained language models encode correlations between social groups and traits, like associating the group with the group. |
| Approach: | They adapt the Agency-Belief-Communion (ABC) stereotype model to a language model and introduce the sensitivity test (SeT) to measure stereotypical associations. |
| Outcome: | The proposed framework is used to measure stereotyping of intersectional identities in language models. |
SHAPE: Unifying Safety, Helpfulness and Pedagogy for Educational LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing educational LLMs are vulnerable to pedagogical jailbreaks where students use answer-inducing prompts to elicit solutions rather than scaffolded instructions. |
| Approach: | They propose a graph-augmented tutoring pipeline that infers prerequisite concepts from queries and identifies mastery gaps. |
| Outcome: | The proposed method improves safety under two pedagogical jailbreak scenarios while maintaining near-ceiling helpfulness under the same evaluation protocol. |
CLaRE-ty Amid Chaos: Quantifying Representational Entanglement to Predict Ripple Effects in LLM Editing (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are outdated or incorrect over time due to unintended ripple effects that propagate even to the hidden space. |
| Approach: | They propose a lightweight representation-level technique to identify where ripple effects may occur by detecting entanglement between facts using forward activations from a single intermediate layer. |
| Outcome: | The proposed method achieves 62.2% improvement in Spearman correlation with ripple effects while being 2.74 faster and using 2.85 less peak GPU memory. |