method · first seen 4 Oct 2026
logit lens
Also: tuned lens
Technique projecting intermediate transformer layer states into the output vocabulary space to inspect predictions across depth.
Mentions per week
PapersCodeDiscussion
Rank 168 on 16 Nov 2026 · 9 mentions in 28 days · 9 independent groups · papers → code · first seen 6 weeks ago · acceleration -0.04
Earliest sources
- Belief-Trajectory Energy: Measuring the Path to a PredictionPapers · 4 Oct 2026 · Jiahao Ying, Wei Tang, Boxian Ai, Yaoning Wang, Haotian Chen, Wenhe Sun, Caijun…
- A Sense of Proprioception in Language ModelsPapers · 12 Oct 2026 · Jonas Geiping
- DANGERLENS: Probing Dangerous Representations Hidden by Safety AlignmentPapers · 14 Oct 2026 · Tianzhuo Yang, Shaoheng Yan, Lin Lu, Jiawei Chen, Kaile Wang, Fanyu Meng, Minglu…
- Lexical Reference Frames: Interpreting LM Residual Streams in Native SemanticsPapers · 15 Oct 2026 · Jian Gu, Hongyu Zhang, Chunyang Chen, Aldeida Aleti
- Foreword: Pre-Generation Prediction and Characterisation of LLM Test Failures using the Jacobian LensPapers · 20 Oct 2026 · Somin Kim, Robert Feldt, Shin Yoo
- Angular Supervision for Visual Object Grounding in Autoregressive Vision-Language ModelsPapers · 23 Oct 2026 · Parsa Esmaeilkhani, Longin Jan Latecki
Latest sources
- manuelgscom/inside-llmCode · 15 Nov 2026 · manuelgscom · 92 stars
- Dancing Tokens on the Vocabulary Network from LLM GenerationPapers · 13 Nov 2026 · Yuan-Chao Hu
- MicroLens: Read, Trace, and Rewire the Residual Stream with Two LensesPapers · 9 Nov 2026 · Nika Zahedi, Fangyi Zhu, Anna Wu
- A Layer-wise Study of How Style Rewriting Breaks the Calibration of Zero-Shot LLM Fake-News DetectorsPapers · 8 Nov 2026 · Prabhjot Singh, Praney Goyal, Raksha Sharma
- Forcing a Letter Suppresses Refusal: Mechanistic Evidence from WMDPPapers · 5 Nov 2026 · Stepan Epifantsev, Anna Antipova, Oleg Chernikov, Leon Strelkov, Alexey Dontsov
- stdlogits/neuron-lensCode · 30 Oct 2026 · stdlogits · 53 stars
- The Geometry of Emotion Representations Across Scale, Depth, and Instruction TuningPapers · 26 Oct 2026 · Adam Hollowell
- Angular Supervision for Visual Object Grounding in Autoregressive Vision-Language ModelsPapers · 23 Oct 2026 · Parsa Esmaeilkhani, Longin Jan Latecki
- Foreword: Pre-Generation Prediction and Characterisation of LLM Test Failures using the Jacobian LensPapers · 20 Oct 2026 · Somin Kim, Robert Feldt, Shin Yoo
- Lexical Reference Frames: Interpreting LM Residual Streams in Native SemanticsPapers · 15 Oct 2026 · Jian Gu, Hongyu Zhang, Chunyang Chen, Aldeida Aleti
- DANGERLENS: Probing Dangerous Representations Hidden by Safety AlignmentPapers · 14 Oct 2026 · Tianzhuo Yang, Shaoheng Yan, Lin Lu, Jiawei Chen, Kaile Wang, Fanyu Meng, Minglu…
- A Sense of Proprioception in Language ModelsPapers · 12 Oct 2026 · Jonas Geiping