Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training

Abstract:Large language models offer broad capabilities, but adapting them to evolving domains, tools, and requirements often entails repeated post-training. Autonomous systems automate parts of this process by proposing updates, training candidates, and using evaluation feedback to select subsequent proposals. As evidence accumulates, a central problem emerges: which past update evidence remains actionable after subsequent training has changed the parent model? An update's effect depends on its parent, data, and training stage. Treating past success as context-free permission can waste compute. If the resulting child is promoted, it can also degrade the subsequent training trajectory. We formulate this problem as conditional experience transfer and introduce Boundary-Calibrated Intervention Transfer (BCIT), a method that authorizes experience reuse before weight-changing training. BCIT binds an observed effect to its source context, checks applicability conditions, vetoes candidates with named hard conflicts, and obtains current-state evidence through a bounded training trial when needed. Fully trained candidates still face a shared adoption rule, and only observed events extend memory. On one 4B model adapted across finance reasoning, text-to-SQL, and function calling, candidate updates exhibit heterogeneous target and retention effects across the evaluated contexts. Under matched candidates, evidence, and compute, BCIT authorizes fewer harmful updates and attains higher equal-budget final-model quality than the evaluated alternatives. These results support treating experience authorization as a distinct problem in autonomous post-training.

Artificial Intelligence (cs.AI)

arXiv:2608.26730 [cs.AI]

(or arXiv:2608.26730v1 [cs.AI] for this version)

https://doi.org/10.48550/arXiv.2608.26730 Focus to learn more arXiv-issued DOI via DataCite (pending registration)

reproducibility/README.md

reproducibility/figure_generation/data/README.md

reproducibility/figure_generation/data/equal_budget_trajectories.csv

reproducibility/figure_generation/data/shortfull_24.csv

reproducibility/figure_generation/data/transfer_heterogeneity_24.csv

reproducibility/figure_generation/make_experiment_figures.py

reproducibility/figure_generation/requirements.txt

reproducibility/p0_author_verified_endpoints_20260725.csv

reproducibility/p0_author_verified_process_20260725.csv

reproducibility/protocol_examples/pilot24_protocol/FREEZE_HASHES.json

reproducibility/protocol_examples/pilot24_protocol/candidate_execution_manifest_v1.jsonl

reproducibility/protocol_examples/pilot24_protocol/controller_eval_contract_v1.json

reproducibility/protocol_examples/pilot24_protocol/main_manifest.json

reproducibility/protocol_examples/pilot24_protocol/main_semantics_v1.json

reproducibility/protocol_examples/pilot24_protocol/proposal_prompt_main_v1.txt

reproducibility/protocol_examples/pilot24_protocol/proposal_schema.json

reproducibility/protocol_examples/pilot24_protocol/training_config.json

reproducibility/protocol_examples/proposal_fixture.json

reproducibility/recompute_p0_results.py

reproducibility/requirements.txt

reproducibility/supplement_tables/additive_veto_seeds.csv

reproducibility/supplement_tables/bcit_seeds.csv

reproducibility/supplement_tables/flat_additive_seeds.csv

reproducibility/supplement_tables/shortfull24.csv

reproducibility/supplement_tables/validate_all_seeds.csv

(20 additional files not shown) You must enabled JavaScript to view entire file list.

arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.