Can We Make an LLM Dumber, or Is It Just Playing Dumb? An Empirical Study of Believable Performance Throttling
cs.CL
Submitted: 2026-11-11
Updated: 2026-11-11
Terminology
Sources
- Below-Chance Blindness: Prompted Underperformance in Small LLMs Produces Positional Bias Rather than Answer Avoidance
- Option-Order Randomisation Reveals a Distributional Position Attractor in Prompted Sandbagging
- SCALPEL: Selective Capability Ablation via Low-rank Parameter Editing for Large Language Model Interpretability Analysis
- Distilling the Knowledge in a Neural Network
- LoRA: Low-Rank Adaptation of Large Language Models
- Sleeper Agents: Training Deceptive LLMs that Persist Through Safety Training
- Editing Models with Task Arithmetic
- Frontier Models are Capable of In-context Scheming
- tinyBenchmarks: evaluating LLMs with fewer examples
- Leveraging Large Language Models for Multiple Choice Question Answering
- Model evaluation for extreme risks
- Auditing Games for Sandbagging
- Noise Injection Reveals Hidden Capabilities of Sandbagging Language Models
- Steering Language Models With Activation Engineering
- Representation Engineering: A Top-Down Approach to AI Transparency
Related papers
- ALARM: Audio-Language Alignment for Reasoning Models
- ThinkSum: Probabilistic reasoning over sets using large language models
- Coherence boosting: When your pretrained language model is not paying enough attention
- From Facts to Insights: A Persona-Driven Dual Memory Framework and Dataset for Role-Playing Agents
- Improving General Role-Playing Agents via Psychology-Grounded Reasoning and Role-Aware Policy Optimization
- The Art of Asking: Multilingual Prompt Optimization for Synthetic Data