CORTEXA
← Browse
arxivcs.LGcs.AIstat.ML2026-07-13

Are we Merging the Right Models? Impact of Expert Training Duration on Model Merging for LLMs

Nikita Kozodoi, Zainab Afolabi, Jack Butler

Multi-task model merging combines separately trained expert models into a single model that handles all tasks without co-training. Standard practice merges experts at their optimal validation loss. We challenge this convention by systematically studying how training duration of domain experts affects the quality of the merged model. We fine-tune experts on five domains (Math, Code, Instruction Following, Multilingual, and Safety) across three model sizes (Qwen 3.5 0.8B, 2B, and 4B), saving checkpoints from 25% to 500% of the optimal training steps and evaluating five merging methods at each duration. Our findings reveal a striking method-dependent pattern: simple averaging degrades sharply with overfitting, while sparsification-based methods achieve their best performance well past the validation optimum. We formalize this through bias-variance decomposition analysis, drawing a parallel to random forests where averaging benefits from high-variance individual learners. These results suggest that training duration and merging method should be chosen jointly rather than independently.

View free PDFSource page

Related papers

arxivcs.LGcs.AIstat.ML2026-07-31

When Does On-Policy Interaction Help? Representational Tradeoffs in Value-Based Imitation Learning

Luca Viano, Antoine Moulin, Audrey Huang, Volkan Cevher, Philip Amortila, Dylan J. Foster

Imitation learning (IL)---training an agent to replicate expert behavior from demonstrations---underpins applications from robotics to language model training. Standard approaches such as Behavior Cloning (BC) are known to suffer from compounding errors and performance plateaus,…

View free PDFSource page
arxivmath.OCcs.AIcs.LGstat.ML2026-07-24

Explicit Iteration Complexity of Exact Data-Driven Inverse Optimization for Integer Linear Programs

Akira Kitaoka

A data-driven inverse optimization problem (DDIOP) is the problem of estimating the objective-function parameters (weights) that explain observed optimal-solution data, and it arises in many applications, including integer linear programming (ILP). It is known that, by applying g…

View free PDFSource page
arxivstat.MLcs.AIcs.LGecon.EM2026-07-24

CausalForge: A Formally Grounded, Self-Improving Agentic Framework for Automated Research in Causal Inference

Jiyuan Tan, Vasilis Syrgkanis

Automating theoretical research is constrained not only by the generation of candidate results, but also by their reliable evaluation. A common approach is to close the research loop with a large language model (LLM) reviewer. However, such reviewers remain empirically unreliable…

View free PDFSource page
arxiveess.AScs.AIcs.LGcs.SD2026-07-31

Stable Autoregressive Speech Generation with Low-Frame-Rate High-Dimensional Continuous Tokens

Yi Luo, Rongzhi Gu, Jixun Yao

Balancing sequence length, representational capacity, and long-horizon stability is a central problem in autoregressive (AR) speech and audio generation. Representations with higher frame rates or greater capacity can preserve more signal detail, but they also make streaming gene…

View free PDFSource page
arxivcs.LGcs.AIcs.CL2026-07-24

Teaching LLMs to Self-Evolve: Cultivating Core Meta-Skills with Reinforcement Learning

Shujin Wu, Cheng Qian, Xiusi Chen, Heng Ji

Test-time scaling through iterative self-evolution with environment feedback, as demonstrated by AlphaEvolve, shows remarkable performance gains. We hypothesize that the success of such evolution frameworks hinges on meta-skills, such as self-reflection with environment feedback,…

View free PDFSource page