trends.barnas.net

method · first seen 20 Oct 2026

reward shaping for budget-constrained agents

Also: length penalties, budget-aware reward design

Training reward functions that penalise resource overspending so LLM agents learn to comply with inference budgets.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

AppWorldsigmoid overspending penaltychain-of-thoughtreward shaping