Rock Tokens in on-policy distillation persist at high loss, account for up to 18% of outputs, absorb large gradient norms, but add negligible value to reasoning performance.
In: 2026 9th International Symposium on Big Data and Applied Statistics (ISBDAS)
3 Pith papers cite this work. Polarity classification is still indexing.
3
Pith papers citing it
years
2026 3verdicts
UNVERDICTED 3representative citing papers
MOSAIC combines frozen-LLM semantic embeddings with hierarchical consistency objectives to report up to 3.4% AUC gains on knowledge-tracing benchmarks including a new MOOC dataset.
Introduces failure-aware observability framework for diagnosing wasted computation in multi-agent LLM systems and evaluates it on 165 GAIA traces showing common operational failures.
citing papers explorer
-
MOSAIC: Orchestrating Collaborative Knowledge Tracing with Hierarchical Semantic Alignment
MOSAIC combines frozen-LLM semantic embeddings with hierarchical consistency objectives to report up to 3.4% AUC gains on knowledge-tracing benchmarks including a new MOOC dataset.