AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

Signal in the Noise: Polysemantic Interference Transfers and Predicts Cross-Model Influence

Bofan Gong, Shiyang Lai, James Evans, Dawn Song

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
prepend high-interference tokens from small models instead of random tokensPythia-70M · Prompt injection, Locations targetsuccess rate (%): 35.06% -> 36.93%Improved
prepend high-interference tokens instead of random tokensLlama-3.1-70B-Instruct · Prompt injection, Number targetsuccess rate (%): 7.09% -> 8.85%Improved
prepend high-interference tokens instead of random tokensLlama-3.1-8B-Instruct · Prompt injection, Locations targetsuccess rate (%): 18.24% -> 20.78%Improved
prepend high-interference tokens instead of random tokensGemma-2-9B-Instruct · Prompt injection, Science targetsuccess rate (%): 20.15% -> 20.40%No clear effect
prepend high-interference tokens instead of random tokensGPT-2-Small · Prompt injection, Number targetsuccess rate (%): 34.71% -> 30.33%No clear effect
prepend high-interference tokens instead of random tokensPythia-70M · Prompt injection, Science targetsuccess rate (%): 28.58% -> 23.13%No clear effect
prepend low-interference tokens instead of random tokensLlama-3.1-70B-Instruct · Prompt injection, Locations targetsuccess rate (%): 24.48% -> 23.09%No clear effect
use token-gradient steering instead of SAE feature direction steeringPythia-70M and GPT-2-Small · next-token steeringeffect size: ~10x largerImproved
use prompt injection instead of direct latent-space steeringPythia-70M and GPT-2-Small · next-token steeringeffect size: ~10% levelHurt
choose high-interference over low-interference matched pair, judged by LLMGPT-5-mini · high-interference vs low-similarity-matched pairspair choice rate: 64.3%Improved
amplify activations of high-polysemanticity super-neurons instead of masking them (ablation)GPT-2-Small and Pythia-70M · neuron interventionweighted cosine change: masking less influence; amplifying exponentially greaterMixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy