GPT-2 small solves indirect object identification via a circuit of 26 attention heads organized into seven functional classes discovered through causal interventions.
A ttention is not E xplanation
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
citation-role summary
background 2
method 1
citation-polarity summary
polarities
background 3representative citing papers
SGC-RML creates an 8D symptom atlas from multimodal PD data and integrates conformal calibration to deliver reliable, rejectable longitudinal assessments.
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.
RETRO matches GPT-3 and Jurassic-1 performance on the Pile benchmark using 25 times fewer parameters by conditioning on retrieved chunks from a 2-trillion-token database.
LLMs generate adequate counterspeech for co-occurring hate and misinformation in 40% of cases, with a mixed knowledge strategy from fact-checkers and NGOs proving most effective after expert revision.
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.
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
-
Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 small
GPT-2 small solves indirect object identification via a circuit of 26 attention heads organized into seven functional classes discovered through causal interventions.
-
SGC-RML: A reliable and interpretable longitudinal assessment for PD in real-world DNS
SGC-RML creates an 8D symptom atlas from multimodal PD data and integrates conformal calibration to deliver reliable, rejectable longitudinal assessments.
-
MASPrism: Lightweight Failure Attribution for Multi-Agent Systems Using Prefill-Stage Signals
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.
-
Hidden in the Multiplicative Interaction: Uncovering Fragility in Multimodal Contrastive Learning
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.
-
Improving language models by retrieving from trillions of tokens
RETRO matches GPT-3 and Jurassic-1 performance on the Pile benchmark using 25 times fewer parameters by conditioning on retrieved chunks from a 2-trillion-token database.
-
Assisted Counterspeech Writing at the Crossroads of Hate Speech and Misinformation
LLMs generate adequate counterspeech for co-occurring hate and misinformation in 40% of cases, with a mixed knowledge strategy from fact-checkers and NGOs proving most effective after expert revision.
-
ORBIT: Learning Gene Program Co-Activation Structure for Cell-Type-Stratified Pathway Rewiring Analysis in Single-Cell Transcriptomics
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.
-
Matched-Learning-Rate Analysis of Attention Drift and Transfer Retention in Fine-Tuned CLIP
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.