merging
Datasets
All datasets matching “merging”Mergingaim-activation-informed-merging
AIM: does activation-informed merging change what makes a merge work?
Headline
AIM does exactly what it claims, the targeting is what makes it work — and it changes
nothing about what predicts a good merge.
AIM is exactly what it says on the tin, and that is verifiable from public artefacts alone.
The published with-AIM checkpoints are recovered, to R² = 0.9992, as a closed-form
per-input-channel shrinkage of their baseline twins toward the base model, with ω̂ =… See the full description on the dataset page: https://huggingface.co/datasets/Cross-Mergeability/aim-activation-informed-merging.peft_merging_datamechanism-merging-exp089-handoff
Mechanism Merging Exp089 analysis handoff
这是 weekly_experiment_report_20260812.md 对应版本的公开研究交付仓库。它保存报告第 8、10、11 章
所用的模型、训练数据、逐样本评测轨迹和分析实物；代码、可直接浏览的报告与图片位于私有 GitHub
仓库 xlxcomputer/mechanism-merging-exp089-repro。
四个分析输入模型
目录
角色
冻结点
训练成本口径
checkpoints/anchor_mixed_stage1_step10014/huggingface/
三种方法共同且唯一的 mixed Stage-1 anchor
step 10014
共同成本，不进入方法横轴
checkpoints/mix-rl_step1036/huggingface/
Mix-RL final
step 1036
online FLOPs 100210702010088587264… See the full description on the dataset page: https://huggingface.co/datasets/2041Xu/mechanism-merging-exp089-handoff.livecodebench-merging-leaderboard
LiveCodeBench v6 Evaluation Leaderboard
Evaluation results for cross-capability merging of OLMo-3 and OLMo-3.1 RL-Zero models on 454 coding problems.
Evaluation
We followed the evaluation guidelines and prompts from OLMo 3. Best effort was made to ensure reported numbers are as accurate as possible.
Code: pmahdavi/modal-eval
Leaderboard
Model
pass@4
pass@1
Loop Rate
Qwen/Qwen3-4B-Thinking-2507
54.6%
45.4%
0.4%
pmahdavi/Olmo-3-7B-Think-Math-Code… See the full description on the dataset page: https://huggingface.co/datasets/pmahdavi/livecodebench-merging-leaderboard.details_Cartinoe5930__MoE-Merging
Dataset Card for Evaluation run of Cartinoe5930/MoE-Merging
Dataset automatically created during the evaluation run of model Cartinoe5930/MoE-Merging on the Open LLM Leaderboard.
The dataset is composed of 63 configuration, each one coresponding to one of the evaluated task.
The dataset has been created from 1 run(s). Each run can be found as a specific split in each configuration, the split being named using the timestamp of the run.The "train" split is always pointing to the… See the full description on the dataset page: https://huggingface.co/datasets/open-llm-leaderboard-old/details_Cartinoe5930__MoE-Merging.