Target-Token Asymmetry Can Masquerade As Depth Specialization

cs.CL, cs.LG

Submitted: 2026-10-14

Updated: 2026-10-14

Comments: 48 pages, 7 figures

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Sources

• Eliciting Latent Predictions from Transformers with the Tuned Lens
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• TinyStories: How Small Can Language Models Be and Still Speak Coherent English?
• Gemma 2: Improving Open Language Models at a Practical Size
• The Llama 3 Herd of Models
• CodeSearchNet Challenge: Evaluating the State of Semantic Code Search
• Locating and Editing Factual Associations in GPT
• Pointer Sentinel Mixture Models
• Qwen2.5 Technical Report
• Qwen3 Technical Report
• Are Emergent Abilities of Large Language Models a Mirage?
• Hermes 3 Technical Report
• Towards Best Practices of Activation Patching in Language Models: Metrics and Methods

Related papers

• Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
• Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
• Subliminal Steering: Stronger Encoding of Hidden Signals
• MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
• The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
• Untangling the Mechanisms of Misleading Context in Medical Question Answering