EMNLP paper probes why some LLM biases survive rewording while others don't
radamihalcea · x · 2026-10-27
A paper titled "Deep and shallow biases in language models", presented at EMNLP 2026 in Budapest (poster 1152-MAIN), examines the robustness of LLM biases: ask a model for a random number and you get 42, but reword the prompt and it says 3; ask for a random butterfly and you get Monarch, which survives rewording. The authors investigate why some biases vanish under paraphrase while others persist, with a demo video accompanying the work.
More from Research
- Kolab Benchmark Uses Pro Screen Recordings to Show AI Agents Overestimate Collaboration Quality — minnesotanlp · 2026-10-27
- One BBQ Example Pushes LLM Fairness Scores Past 90%, EMNLP Paper Questions Benchmarks — radamihalcea · 2026-10-27
- Real2Sem2Real turns one phone video into 2,000+ robot demos, matching 150 teleop demos — Scobleizer · 2026-10-27
- Understanding MuonH (Hyperball) through weight norm geometry of scale-invariant networks — jasondeanlee · 2026-10-27
- Sail equivalence proof and riscv-coq cross-checks uncover real RISC-V spec bugs — xennygrimmato_ · 2026-10-27
- Kokotajlo challenges CoT monitorability evals: what if the model didn't refuse? — dfrsrchtwts · 2026-10-27