Δτ-Bench tests agents as tool ecosystems change
The strongest evaluated model or harness reached only 42.3% success on 1,588 tasks across six ecosystem states.
Stories per day · 30 days
2 RSS
Relatedagent-benchmark1credit-assignment1ecosystem-drift1hotpotqa1locomo1multi-agent1steward1tool-use1
8 stories, newest first · one card per story
Stories tagged #skill-learning
GRSE credits multi-agent skills by contrasting repeated runs
On HotpotQA and LoCoMo it beat MASkills on most categories, with the largest gains on multi-hop questions.
Socratic Skill Learning interleaves teaching with agent practice
Fixed-Window SSL scored 0.801 utility on 155 cases, above no-Skill, Trace2Skill, and SKILL-KD baselines.
RoboRSI: Stable, efficient, and reusable robot self-evolution in complex real-world environments
RoboRSI lets a mobile manipulator reuse validated skills, beating baselines by up to 11 points.
UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy
UniSkill lets one policy act and propose skill edits using contrastive feedback without extra rollouts.
Skill2Real: Agentic Skill Learning for Zero-Shot Sim-to-Real Robot Manipulation
Skill2Real learns sim-to-real robot manipulation skills via a Proposer-Verifier-Governor loop, achieving 78.75% task completion on real robots without fine-tuning.
Skill-Space Shooting for Autonomous Robot Policy Improvement
Paper introduces skill-space shooting, using foundation model guidance over reusable skills to turn robot failure corrections into autonomous policy improvement.
Making Agents More Consistent: Skills Should Form Habits for Repeat Tasks
Paper proposes skill habit formation so agents cache deterministic skills from execution history, fixing 38-74% answer inconsistency and cutting tokens 14-56%.