Continual-ARC: Measuring Skill Acquisition over a Working Lifetime
Daniel Lenton, Yusha Arif, Haris Mahmood
Continual-ARC: Measuring Skill Acquisition over a Working Lifetime: 2 upvotes on Hugging Face Daily Papers, #37 of 40 papers on 2026-11-16. Day-by-day upvote history. It lost 8 votes when the Hub removed votes in bulk.
Upvotes over time
It had 10 upvotes by its first day, 10 upvotes by the next day, 2 upvotes by day 3, 2 upvotes by day 7, and has 2 now. It ranked #113 of 146 papers in its week and #444 of 520 in its month. Its code repository has 38 GitHub stars.
Agent harnesses have emerged as a promising mechanism for continual learning, alongside test-time parametric learning. However, the current benchmarks have several limitations: they generally focus on a single applied domain, serve each instance only once, and assume a single family of learner, so memory-based and weight-based systems are rarely compared on the same scale. We propose Continual-ARC, a lifelong stream built from the three ARC-AGI editions, whose tasks need no domain knowledge. The same tasks return after dormant gaps with fresh instances, while new tasks keep arriving. On each instance, the learner may pay for demonstrations, submit answers, or act inside the harness, with retries permitted on incorrect submissions, which are paid for. Each instance is scored based on the share of this feedback budget consumed. Since a learner may keep anything between instances, memory-based and weight-based systems are scored on the same scale. After the stream, the learner is frozen and tested on fresh instances of the tasks it met, which measures memorization, and on unseen tasks, which measures generalization. We run an agent harness with frozen weights, with and without persistence, on Qwen3.5-9B and GLM-5.3 Flash, comparing them with two prompting floors, a program-induction method and Qwen3.5-9B's weights trained to imitate GLM-5.3 Flash. Persistence in the harness generally does not improve memorization unless provided with explicit expert guidance. Most harness-based methods still learn less than a floor that simply caches each task's demonstrations, across all models we tested. No harness-based method gains on unseen tasks, whereas weight-based learning shows substantial gains. This opens the door to future weight-based continual learning that could teach models to drive harnesses in new domains.
Paper page on Hugging Face · arXiv
Data: hysts-bot-data/daily-papers-stats and the Daily Papers API. Open data: tardellirs/paper-pulse-data. Sister project: Model Pulse, the download history of every model on the Hub.