← Back to all news
arXiv Machine Learning

On Emergent Capabilities and Model Merging

The paper investigates how model merging—an arithmetic operation on the weights of fine‑tuned checkpoints and adapters—affects emergent capabilities, which are behaviors not explicitly trained for. Across two testbeds and three model families, the authors find that merging preserves shared emergent capabilities, cannot create superadditive capabilities, and tends to dilute a capability that is present in only one parent. Thus, emergent behaviors do not compose in the same way as trained capabilities.

Related stories

arXiv Machine Learning
Sep 11

A Fragility Spectrum for Recursive Language-Model Training

The paper investigates how recursive contamination—retraining language models on their own generated text—affects output diversity across 13 publicly released checkpoints. Using a fixed contamination protocol over five generations, the authors find a wide spread in 4‑gram diversity (0.187 to 0.940), indicating that some models collapse into repetitive fragments while others remain largely unaffected. The study shows that a model’s susceptibility to collapse is an intrinsic property of the checkpoint, not predicted by parameter scale or static indicators, and that simple interventions such as tightening top‑p sampling can significantly slow or halt collapse.

By Yangze Liu, Zhongyi Han
More like this →
arXiv AI
Aug 7

When Self-Evolution Backfires: Pre-Commit Gating against Skill Contamination in LLM Agents

arXiv:2608. 05810v1 Announce Type: new Abstract: Self-evolving agents accumulate capability by distilling reusable skills from their execution trajectories, but we find this process is not monotonic: past a critical pool size, newly added skills degrade performance instead of improving it.

By Linfang Shang, Ming Xu, Yiding Sun, Tianle Xia, Lingxiang Hu, Lan Xu, Ning Zheng
llmsagentsbenchmarks
More like this →
arXiv AI
Sep 17

EvoUndo: Recoverability-Constrained Self-Evolution for LLM Agent Harnesses

EvoUndo is a framework that enables large language model agents to self‑evolve—modifying prompts, tools, and execution harnesses—while ensuring that these changes can be reliably reversed across different states. The study evaluates EvoUndo on 600 unseen one‑shot self‑evolution tasks, finding that 197 capability‑improving mutations fail recoverability checks. By extending the recovery language and adding exact state‑address diagnostics, the framework recovers up to 191 out of 197 failures, demonstrating that robust self‑evolution requires co‑designing verification, grounding, witness semantics, and recovery expressivity.

By Tanmay Sah, Dolly Sah, Harshul Jain, Tanya Sah
llmsagents
More like this →
arXiv Computation and Language
Aug 31

Fidelity Is Not Enough: Dispatch-Level Instrumentation for Agentic Datasheet Extraction

The paper reports that a model can pass fidelity checks—verifying that extracted values match the source—without actually opening a datasheet, due to a hidden constraint that disables tool use. To address this, the authors log every tool call in an agentic benchmark and develop two instruments: a rule‑based failure‑attribution classifier and a silent‑failure detector that flags runs based solely on which tools were invoked. While the detector shows low false positives on clean extractions and recovers all planted faults, its recall against correct tool usage but incorrect answers remains unmeasured, and a partial causal chamber confirms only a subset of claims, highlighting limitations in physical verification.

By Qing Ye, Meng-Hsuan Lin
llmsagentsbenchmarks
More like this →