AI Paper Atlas

Research map / Large language models / Instruction and fine-tuning

Bridging Explainability and Embeddings: BEE Aware of Spuriousness

Cristian Daniel Paduraru, Antonio Barbalau, Radu Filipescu, Andrei Liviu Nicolicioiu, Elena Burceanu

ICLR 2026 · Large language models › Instruction and fine-tuning · approach: Fine-tuning for reasoning data · 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
regularize class weights to be equidistant from BEE-found SCslinear probe, Waterbirds without counterexamples · Waterbirds (no counterexamples)worst-group accuracy: 43.2 -> 57.9Improved
regularize class weights to be equidistant from BEE-found SCslinear probe, CivilComments without counterexamples · CivilComments (no counterexamples)worst-group accuracy: 18.6 -> 31.3Improved
regularize class weights to be equidistant from BEE-found SCslinear probe, CelebA without counterexamples · CelebA (no counterexamples)worst-group accuracy: 9.6 -> 10.4Improved
use BEE-identified SCs instead of random class-neutral concepts for regularization (ablation)SC regularization, Waterbirds · Waterbirds (no counterexamples)worst-group accuracy: 46.6 -> 57.9Improved
use BEE-identified SCs instead of random class-neutral concepts for regularization (ablation)SC regularization, CivilComments · CivilComments (no counterexamples)worst-group accuracy: 19.1 -> 31.3Improved
use BEE SCs instead of Lg's SCs for regularization (ablation)SC regularization, CelebA · CelebA (no counterexamples)worst-group accuracy: 8.3 -> 10.4Improved
train with GroupDRO instead of ERM when no minority groups existlinear probe, Waterbirds · Waterbirds (no counterexamples)worst-group accuracy: 43.2 -> 38.9Hurt
add BEE-found spurious concept firemen to peafowl imageCLIP ViT-L/14 fine-tuned on ImageNet · FLUX-generated peafowl imagescorrect class rate (%): 100.0 -> 5.3Hurt
add BEE-found spurious concept reading a newspaper to Mexican hairless dog imageCLIP ViT-L/14 fine-tuned on ImageNet · FLUX-generated Mexican hairless dog imagescorrect class rate (%): 47.5 -> 0.9Hurt
add BEE-found spurious concept shrimp to Bernese Mountain Dog imageCLIP ViT-L/14 fine-tuned on ImageNet · FLUX-generated Bernese Mountain Dog imagescorrect class rate (%): 99.8 -> 10.6Hurt
add firemen to peafowl prompt, evaluated on alexnetalexnet trained on ImageNet-1k · Generated peafowl + firemen (transfer)accuracy (%): 100.0 -> 4.6Hurt
add firemen to peafowl prompt, evaluated on swin_bswin_b trained on ImageNet-1k · Generated peafowl + firemen (transfer)accuracy (%): 100.0 -> 81.5Hurt

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