AI Paper Atlas

Research map / Architectures and efficiency / RNNs and normalization

FlowNIB: An Information Bottleneck Analysis of Bidirectional vs. Unidirectional Language Models

Md Kowsher, Nusrat Jahan Prottasha, Shiyun Xu, Shetu Mohanto, Niloofar Yousefi, Ozlem Garibay, Chen Chen

ICLR 2026 · Architectures and efficiency › RNNs and normalization · approach: Layer-wise analysis of language models · 0 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
use masked-token prediction inference instead of poolingDeBERTa-v3-Base · 9 classification tasks avgaccuracy (avg): 77.90 -> 81.52Improved
use masked-token prediction inference instead of poolingDeBERTa-v3-Large · 9 classification tasks avgaccuracy (avg): 80.96 -> 84.73Improved
use masked-token prediction inference instead of poolingRoBERTa-Base · 9 classification tasks avgaccuracy (avg): 76.53 -> 79.95Improved
use masked-token prediction inference instead of poolingRoBERTa-Large · 9 classification tasks avgaccuracy (avg): 80.14 -> 83.95Improved
use masked-token prediction inference instead of poolingModernBERT-Base · 9 classification tasks avgaccuracy (avg): 76.74 -> 79.73Improved
use masked-token prediction inference instead of poolingModernBERT-Large · 9 classification tasks avgaccuracy (avg): 80.35 -> 83.84Improved
use generation-based prediction inference instead of poolingGPT-2 Medium · 9 classification tasks avgaccuracy (avg): 71.02 -> 72.04Improved
use generation-based prediction inference instead of poolingGPT-2 Large · 9 classification tasks avgaccuracy (avg): 71.26 -> 72.07Improved
use generation-based prediction inference instead of poolingSmolLM2-135M · 9 classification tasks avgaccuracy (avg): 71.37 -> 72.82Improved
use generation-based prediction inference instead of poolingSmolLM2-360M · 9 classification tasks avgaccuracy (avg): 72.95 -> 74.40Improved
use generation-based prediction inference instead of poolingMobileLLM-125M · 9 classification tasks avgaccuracy (avg): 70.48 -> 71.92Improved
use generation-based prediction inference instead of poolingMobileLLM-350M · 9 classification tasks avgaccuracy (avg): 71.89 -> 73.73Improved

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