Dispersion loss (LM-Dispersion)
- Dispersion loss counteracts embedding condensation and improves generalization in small language models *Equal contribution ICML 2026 One-liner summary What makes LLMs better than small LMs?
Unverified
- Dispersion loss counteracts embedding condensation and improves generalization in small language models *Equal contribution ICML 2026 One-liner summary What makes LLMs better than small LMs?
Sources: Github