Research map / Large language models / Language, safety and interpretability
The Other Mind: How Language Models Exhibit Human Temporal Cognition
Lingyu Li, Yang Yao, Yixu Wang, Chunbo Li, Yan Teng, Yingchun Wang
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 changed | Where | Measured | Effect |
|---|---|---|---|
| scale up model size in the year similarity judgment task | Qwen2.5 and Llama 3 instruct models · year-pair similarity judgment (1525-2524) | fit to reference-log-linear distance | Improved |
| model years as years instead of plain numbers in the prompt (ablation) | similarity judgment task · year vs number control | best-fitting distance (R2) | Mixed |
| use reference-log-linear distance (reference 2025) instead of plain log-linear distance to predict year judgments (ablation) | LLM year similarity · year-pair similarity judgment | R2 | Improved |
| scale up model size for neuron activation pattern | temporal-preferential neurons · mean activation vs year | logarithmic-like compression | Improved |
| increase model depth within layers for log encoding of past years | Qwen2.5-72B-Instruct neurons · layer-wise regression on log distance to 2025 | R2: R2 of 0.756 at layer 71 | Improved |
| compare past vs future years in neural coding | LLM temporal neurons · year similarity judgment | similarity of year pairs | Mixed |
| go from shallow to deep layers for probing year distances | linear probes on residual stream · similarity judgment hidden states | adjusted R2 | Mixed |
| in Qwen models, let reference-distance representation emerge in late layers | Qwen2.5 series · linear probes | adjusted R2 of log-linear distance | Hurt |
| fit reference point as free parameter instead of fixing at 2025 (ablation) | reference-log-linear distance · year similarity | interpretability of fit | Hurt |
| analyze embedding models of year strings | Qwen3-Embedding-8B, text-embedding-3-large, Gemini-embedding-001 · MDS of year embeddings | structure of year space | No clear effect |
Closest papers by meaning
- Scaling and context steer LLMs along the same computational path as the human brain (NeurIPS 2025 · similarity 0.88)
- Large Language Models as Model Organisms for Human Associative Learning (NeurIPS 2025 · similarity 0.88)
- Language Models Represent Space and Time (ICLR 2024 · similarity 0.88)
- Divergences between Language Models and Human Brains (NeurIPS 2024 · similarity 0.88)
- The Emergence of Abstract Thought in Large Language Models Beyond Any Language (NeurIPS 2025 · similarity 0.87)
- Turning large language models into cognitive models (ICLR 2024 · similarity 0.87)
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.
