When Less is More: The LLM Scaling Paradox in Context Compression

arXiv:2602.09789 · cs.LG · Submitted 2026-02-10 · Read on arXiv

cs.LG

Submitted: 2026-02-10

Updated: 2026-10-30

Terminology

• Decoder — 1× in this paper · explained in 2 episodes
• Faithfulness — 1× in this paper · explained in 2 episodes
• Fidelity — 1× in this paper · explained in 2 episodes
• Reconstruction Error — 1× in this paper · explained in 2 episodes
• Scaling Law — 1× in this paper · explained in 2 episodes
• Scaling Laws — 1× in this paper · explained in 5 episodes
• Semantic Drift — 1× in this paper · explained in 3 episodes

Related papers

• Power Distribution Bridges Sampling, Self-Reward RL, and Self-Distillation
• Train Separately, Merge Together: Modular Post-Training with Mixture-of-Experts
• Democratizing Tool Learning with Environments Fully Simulated by a Free 8B Language Model
• D-STEER: Two-Way Control of DPO-Induced Safety Behavior in Activation Space
• Optimal hypersurface decision trees
• Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation