Larger LLM compressors in lossy setups often yield less faithful context reconstructions due to knowledge overwriting and semantic drift, with mid-sized models outperforming larger ones across 27 tested configurations.
Why do small language models underperform? studying lan- guage model saturation via the softmax bottleneck.arXiv preprint arXiv:2404.07647,
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
When Less is More: The LLM Scaling Paradox in Context Compression
Larger LLM compressors in lossy setups often yield less faithful context reconstructions due to knowledge overwriting and semantic drift, with mid-sized models outperforming larger ones across 27 tested configurations.