AI Paper Atlas

Research map / Large language models / Language, safety and interpretability

Post-Hoc Merging is Not Enough: Many-Shot Model Merging with Loss-Gap Balancing

Kyungjin Im, Miru Kim, Chanin Eom, Minhae Kwon

ICML 2026 · Large language models › Language, safety and interpretability · approach: LLM unlearning and benchmarks · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
switch from post-hoc to many-shot merging for Task Arithmetic (ablation)Task Arithmetic · Llama-3.2-3B multi-tasknormalized performance: 0.706 -> 0.857Improved
switch from post-hoc to many-shot merging for ConsensusTA (ablation)ConsensusTA · Llama-3.2-3B multi-tasknormalized performance: 0.942 -> 0.945No clear effect
switch from post-hoc to many-shot merging for ConsensusTA (ablation)ConsensusTA · Llama-3.2-3B multi-taskmulti-task loss: 1.83 -> 1.49Improved
switch from post-hoc to many-shot merging for TIES (ablation)TIES · Llama-3.2-3B multi-tasknormalized performance: 0.883 -> 0.938Improved
switch from post-hoc to many-shot merging for Task Arithmetic (ablation)Task Arithmetic · Llama-3.1-8B multi-tasknormalized performance: 0.704 -> 0.785Improved
switch from post-hoc to many-shot merging for ConsensusTA (ablation)ConsensusTA · Llama-3.1-8B multi-tasknormalized performance: 0.694 -> 0.898Improved
switch from post-hoc to many-shot merging for DARE (ablation)DARE · Qwen-3-4B multi-tasknormalized performance: 0.986 -> 1.154Improved
switch from post-hoc to many-shot merging for TIES (ablation)TIES · Qwen-3-4B multi-tasknormalized performance: 1.108 -> 1.087Hurt
add loss-gap weighting to many-shot ConsensusTA (METIS)ConsensusTA many-shot · Llama-3.2-3B multi-tasknormalized performance: 0.945 -> 1.015Improved
add loss-gap weighting and consensus masking in many-shot (METIS) vs best post-hocMETIS · Llama-3.2-3B worst-performing tasknormalized performance: 0.641 -> 0.872Improved
remove many-shot merging, loss-gap weighting, or consensus masking from METIS (ablation)METIS · Llama-3.2-3B ablationnormalized performanceImproved
use METIS instead of other merges for knowledge retentionMETIS · CoQA Llama-3.2-3Bexact match: 0.583 -> 0.670Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy