model families
Groups of related models that share a common underlying structure or conceptual framework, such as linear models, tree-based models, or neural networks, facilitating comparative analysis and generalization.
- Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model?
- ErrorTrace: A Black-Box Traceability Mechanism Based on Model Family Error Space
- Guiding LLM Decision-Making with Fairness Reward Models
- Language Models (Mostly) Know When to Stop Reading
- LoTA-QAF: Lossless Ternary Adaptation for Quantization-Aware Fine-Tuning
- Mitigating Forgetting in LLM Fine-Tuning via Low-Perplexity Token Learning
- Quantifying Elicitation of Latent Capabilities in Language Models
- Remarkable Robustness of LLMs: Stages of Inference?