Federated aggregation is the primary driver of performance gains, contributing an average 11.99% improvement¶

Claim · All claims
Evidence · 1 study · design r2 · q2
Currency · newest study 2026 · not yet checked against its sources · not reviewed by a person · recheck by 2029: tool and model accuracy changes quickly

Subclaims¶

q2 i? Removing federation causes the largest degradation in the ablation, with improvements dropping by an average of 11.99%. → Yagna Manasa Boyapati 2026

Evidence¶

Yagna Manasa Boyapati 2026¶

Yagna Manasa Boyapati, Chong Yu, Tianyu Jiang, Justin Zhan. (2026). Privacy-Preserving Heterogeneous Multi-LLM Federated Inference for Cognitive Diagnosis. arXiv preprint. https://arxiv.org/abs/2609.02947 doi:10.48550/arxiv.2609.02947

q2 · i? · design · r2

Ablation study (Table 4) across three datasets; a simple three-model ensemble also confirms model diversity contributes meaningful gains, but the full system outperforms it across all datasets while providing formal (ε,0)-DP guarantees.

"Removing federation causes the largest degradation, with improvements dropping by an average of 11.99%, confirming federated aggregation is the primary driver of performance gains."

Discussion¶

Related Claims¶

• Taxonomy quality drives CLARA performance: degraded taxonomies reduce alignment, and removing SEL causes the largest degradation — related
• Residual correction is essential: removing it while retaining LDP yields MAE worse than the single-LLM baseline across all datasets — related
• Removing the cognitive load module causes the largest ablation performance drop, while state fusion has a smaller effect — related
• Ablations show the semi-Markov controller is the primary driver of behavioral fidelity (removal pushes DKL to 6.81) and Strategist/Executor separation is essential for epistemic fidelity and realism — related
• Pipeline components play complementary roles: hierarchical generation has the largest effect, grounding improves specificity and factual accuracy, and cluster-informed TOC planning improves structure and audience fit — related
• Both RL stages are needed: removing DPO and GRPO training causes the largest performance drop in student simulation — related
• Ablations show each CogEvolution module contributes: removing ICAP perception, structured retrieval, or evolutionary update degrades mistake precision, learning-curve fit, and alignment — related
• Removing the DPO objective reduces examination accuracy by 1.11 percentage points and degrades annotated pedagogical quality — related
• Heterogeneous multi-LLM federated inference improves cognitive diagnosis MAE over single-LLM baselines across three educational benchmarks with minimal privacy cost — possibly the same claim (merge candidate)
• Ablation study shows metadata-enriched embeddings produce the largest accuracy drop, while attention masking, KV caching and dynamic batching primarily support inference efficiency — related
• In the EduAgent framework, removing past cognitive states from memory significantly reduced simulation performance — related
• Ablations show the trustworthiness module contributes most to performance, and the explanation generator cuts human review time by 41% — related
• CLARA's gains come from the interaction of structured representation, constrained annotation, and aggregation, not prompt engineering alone — a broader claim this one bears on