A causal audit with image interventions shows text-only models reach within 5.7 accuracy points of top multimodal VLMs on chest radiography, with some large multimodal models statistically indistinguishable from small text-only baselines.
Attention is not not explanation
12 Pith papers cite this work. Polarity classification is still indexing.
12
Pith papers citing it
citation-role summary
background 1
method 1
citation-polarity summary
years
2026 12representative citing papers
LOCOS scores attention heads via OV-circuit output projection onto answer-token unembedding directions and identifies non-literal retrieval heads whose ablation collapses performance on non-literal benchmarks more than prior literal-copy detectors.
MASPrism attributes failures in multi-agent systems by ranking candidates from prefill-stage NLL and attention signals of a 0.6B SLM, beating baselines by up to 33.41% Top-1 accuracy and proprietary LLMs by up to 89.5% relative improvement while processing traces in 2.66 seconds.
Multimodal contrastive learning using multilinear products is fragile to single bad modalities, and a gated version improves top-1 retrieval accuracy on synthetic and real trimodal data.
PV-TAM uses prompt-side semantics and a bias filter to improve attention-based and IoU localization metrics for vision-language models over answer-side baselines.
CERA fine-tunes a dense retriever with triplet contrastive learning plus attention alignment to human rationales, claiming better retrieval effectiveness and faithfulness on clinical trial reports than Contriever and standard hard-negative baselines.
ORBIT uses an intervention-consistent self-supervised objective in a transformer to infer asymmetric gene program influences from observational scRNA-seq data, recovering Alzheimer's vulnerability patterns and achieving 0.984 macro F1 cell-type classification from 220 pathway scores.
Both humans and LLMs trust content more when labeled human-authored than AI-generated, with LLMs showing denser attention to labels and higher uncertainty under AI labels, mirroring human heuristic patterns.
Empirical Jacobian analysis reveals that token influence in trained language models decays as a power law with distance (exponent ~0.8), a learned property not present in random models.
Profy uses take-level expert-amateur labels on 1083 piano recordings to produce time-aligned highlight scores that correlate with expert review points (r=0.61) on held-out amateur clips.
Rigorous interpretability can function as a principled form of model evaluation if its claims are falsifiable, reproducible, and predictive.
Matched learning-rate experiments show LoRA retains substantially higher zero-shot transfer (45% vs 11% on EuroSAT, 58% vs 9% on Pets) than Full FT in CLIP adaptation.
citing papers explorer
No citing papers match the current filters.