LeafSkill: Efficient Leaf-Level Action Grounding for Recursive LLM Agents

Shen Liu; Ze Zhou; Zhenguo Xu; Shaopu Wang; Yike Gao; Chunlei Wang

cs.AIOct 23, 2026

LeafSkill: Efficient Leaf-Level Action Grounding for Recursive LLM Agents

Authors: Shen Liu, Ze Zhou, Zhenguo Xu, Shaopu Wang, Yike Gao, Chunlei Wang

Organizations: North China Institute of Computer System Engineering, Beijing, China · Beijing University of Aeronautics and Astronautics, Beijing, China · University of Science and Technology of China, Hefei, China · China Information Security Research Institute Co., Ltd., Beijing, China

Abstract

Recursive large language model (LLM) agents solve long-horizon tasks through decomposition, execution, and environmental feedback. Yet after producing an atomic subtask, they often invoke the same large model again to translate a familiar instruction into an executable action. We observe that high-level planning is task-specific, whereas leaf-level action grounding recurs across tasks and can be checked against environment-admitted actions. We introduce LeafSkill, an external reuse layer that retrieves semantically related grounding experience, binds action templates to current objects and admissible actions, and uses a lightweight 1.7B selector to choose a bound action or defer the unchanged subtask to the original agent. Skills remain outside the model parameters, allowing a frozen selector to use post-training additions. Under lexical and syntactic paraphrases, exact matching retrieves none of the reference actions, while semantic retrieval preserves 96.0% and 97.0% candidate coverage; the frozen selector adopts 97.34% of unseen skill types. On an independent confirmation set of 200 Robotouille tasks, LeafSkill completes 137 tasks versus 121 for ReCAP (68.5% vs. 60.5%, exact McNemar p=.0259). On the 106 tasks completed by both, it reduces Qwen3.5-27B tokens by 14.42% and calls by 14.08%. A separate held-out repeat yields a consistent 9.0-point success gain and 18.98% token reduction. Reusable leaf grounding can thus improve task completion while reducing host-model inference without replacing high-level planning.

Figures & tables

Panel of Figure 1 — Figure 1: Host-token composition of ReCAP trajectories used for skill-library construction in Robotouille and ALFWorld.

System	Reused unit	Delegated decision
ExpRAG	Retrieved trajectories	Policy conditions action generation on retrieved experience
GoS	Skill documents and graph neighbors	Host interprets a dependency-aware skill bundle
GraSP	Typed skill DAG	Graph executor composes and repairs multiple nodes
LeafSkill	State-bound action candidates	Local selector chooses an admissible action or abstains

Table 1: Reuse granularity and delegated decisions in representative experience-reuse systems. LeafSkill delegates only a request-local choice over admissible actions.

Figure 2: Overview of LeafSkill. At an eligible leaf task, retrieval recalls templates, binding instantiates admissible actions, and the local selector executes a candidate or abstains to the unchanged host. Source trajectories construct the frozen library and selector supervision offline.

Environment	Adapter boundary	Binding example
Robotouille	Native atomic-leaf flag	pick up [food] from [board] $\rightarrow$ pick up onion1 from board1
ALFWorld	Non-root downward phase; exclude root, upward return, and revisit	check fridge 1 $\rightarrow$ go to fridge 1

Table 2: Environment-specific eligibility and state-bound action instantiation.

Stage	Guard	Outcome
Retrieval	No result above $\gamma$	Return $\bot$
Binding	Missing entity, malformed command, or $a\notin\mathcal{A}_{t}$	Remove candidate; return $\bot$ if empty
Select/resolve	Abstention, unknown ID, or stale mapping	Return $\bot$
Final check	Resolved action not in $\mathcal{A}_{t}$	Return $\bot$

Table 3: Deterministic guards ensure that LeafSkill returns a currently admissible action or leaves execution to the host.

Environment	Split	Size	Role
Robotouille	A	200	Library and selector construction
	B	200	Repeats, ablations, and growth source
	C	200	Independent evaluation
ALFWorld	valid_seen	140	Library and optional adaptation
	valid_unseen	134	Final evaluation

Table 4: Data partitions. Evaluation data never enter library construction or controller adaptation.

Figure 3: ALFWorld cross-run paired-success host-token reduction for four controller routes. Each

3{\times}3

panel crosses ReCAP repeats (rows) with controller repeats (columns), and its title gives the nine-cell mean

\bar{\Delta}_{9}

. Jointly solved counts span 106–109, 112–116, 107–111, and 114–117 in panels (a)–(d).

Figure 4: Robotouille cross-run paired-success host-token reduction. Rows index ReCAP repeats and columns index LeafSkill repeats; each cell reports its pairing-specific reduction and jointly solved count

n

. The title reports the arithmetic mean over all nine cross-run pairings,

\bar{\Delta}_{9}

Metric	$n$	Reduction	95% CI
Tokens	106	14.4%	$[3.0,24.5]$
Calls	106	14.1%	$[3.6,23.3]$

Table 5: Independent Robotouille-C paired-success host-cost reduction.

Evaluation	Run	ReCAP	LeafSkill	$p$	$\Delta$
Robotouille B	R1	55.5	61.5	.126	+6.0
	R2	56.5	65.5	.025	+9.0
	R3	53.0	59.5	.079	+6.5
	Mean	–	–	–	+7.2
Robotouille C	R1	60.5	68.5	.0259	+8.0
ALFWorld unseen	R1	90.3	90.3	1.00	0.0

Table 6: End-to-end success on all scheduled tasks.

\Delta

is the percentage-point difference;

p

is the two-sided exact McNemar result for each run.

Evaluation	Ep.	Calls	Ctrl. tok.	Tok./ep.
Robotouille B R1	200	1,485	–	–
Robotouille B R2–R3	400	3,849	3.98M	9,952
Robotouille C	200	1,355	1.38M	6,900
ALFWorld unseen R1–R3	402	4,636	3.31M	8,236

Table 7: Measured Qwen3-1.7B controller overhead over all scheduled episodes. Tokens include prompts and completions; Robotouille-B R1 predates token instrumentation.

Reuse path	R/B/S	Success ( $p_{2}$ )	Tok. red.
ReCAP R2	–/–/Host	113/200 (–)	–
RawPair-Exact	Ex./Raw/Dir.	121/200 (.3581)	0.20
RawPair-BGE	Sem./Raw/Dir.	127/200 (.0541)	8.08
Canonical-Exact	Ex./St./SFT	134/200 (.0075)	29.60
BGE Top-1	Sem./St./Dir.	124/200 (.2000)	13.08
BGE Top-5 + Base	Sem./St./Base	113/200 (1.000)	27.01

Table 8: End-to-end ablations on Robotouille B. Token reductions average three comparisons with ReCAP R1–R3 (

n=83

–97); success and

p_{2}

compare with R2. R/B/S denotes retrieval/binding/selector.

Diagnostic	$n$	Comparison	Result (%)
Lexical rewrite coverage	200	Exact / Top-1 / Top-5	0.0 / 79.5 / 96.0
Syntactic rewrite coverage	200	Exact / Top-1 / Top-5	0.0 / 70.5 / 97.0
Mixed-rank selection	193	Base / SFT	81.9 / 93.3
All-negative fallback	200	SFT	7.5
Input-scope mixed-rank	193	Local / + planner	93.3 / 90.7

Table 9: Controlled diagnostics: binding coverage (top two rows) and controller accuracy (bottom three).

Library	Keys	Skills	Success	Tok. red.	Call red.
A	413	446	137/200	18.35%	17.70%
A+B	527	588	131/200	25.64%	24.22%

Table 10: Frozen-selector growth on Robotouille C. Cost columns use the same 88 tasks solved by ReCAP and both library variants.

Controller	Training route	Succ. (%)	Use (%)
ReCAP	no controller	89.80	–
Bare 1.7B	none	84.33	79.41
Robot.-only	Robot. SFT	91.29	70.99
ALF-only	ALF seen SFT	86.07	98.02
Sequential	Robot. $\rightarrow$ ALF seen	91.79	98.81

Table 11: ALFWorld success and skill use, averaged over three valid_unseen runs. All routes use the same 228-skill ALFWorld library; cross-run token results appear in Figure 3 .

LLM Agent Skill Retrieval

Large language model agent skill retrieval finds and selects reusable procedures to help agents complete tasks.

See all 102 papers

HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs
Jul 28, 2026Yu Hao, Jinxuan Cai, Qi Zhang +4LLM Agent Skill Learning LLM Agent Skill Retrieval
Beijing University of Posts and Telecommunications · China Mobile Group Shaanxi Co., Ltd · Ant Group
HaReCAP: Habitual-action Grounding for Recursive Large Language Model Agents
Aug 17, 2026Shen Liu, Zhenguo Xu, Shaopu Wang +2Long-Horizon LLM Agents LLM Grounding
North China Institute of Computer System Engineering, Beijing, China · University of Science and Technology of China, Hefei, China · China Information Security Research Institute Co., Ltd., Beijing, China
SCALAR: Learning and Composing Skills through LLM Guided Symbolic Planning and Deep RL Grounding
Mar 10, 2026Renos Zabounidis, Yue Wu, Simon Stepputtis +4LLM-Guided RL LLM Agent Skill Learning
School of Computer Science, Carnegie Mellon University · Department of Mechanical Engineering, Virginia Tech

Jul 28, 2026J/K move · Enter open · S save

HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs

Yu Hao, Jinxuan Cai, Qi Zhang, Yawen Li+3

Beijing University of Posts and Telecommunications · China Mobile Group Shaanxi Co., Ltd · Ant Group

LLM Agent Skill Learning LLM Agent Skill Retrieval Long-Horizon LLM Agents LLM Agent Planning

Open paper AskPDF arXiv

Skills have become an important abstraction for enabling large language model (LLM) agents to reuse past experience in long-horizon interactive tasks. However, existing trajectory-to-skill methods often produce flat collections of high-level textual skills that are stored and retrieved independently, leaving skill relations underutilized and maintaining a gap between high-level skills and executable actions. In this paper, we propose HiSkill, a hierarchical skill graph framework that organizes interaction trajectories into a directed graph with skill nodes, AtomicOp nodes, and typed edges. Specifically, the graph connects reusable high-level skills with executable action templates, while also capturing decomposition, temporal transition, compatibility, support, and recovery relations among them. At inference time, HiSkill retrieves a compact task-relevant subgraph and performs subgraph-guided task execution, where a symbolic task state, an active skill, and the retrieved subgraph guide the LLM agent to switch skills, select AtomicOps, and ground executable actions iteratively. Experiments on three interactive environments show that HiSkill outperforms state-of-the-art baselines while reducing inference token consumption, demonstrating the effectiveness of bridging high-level skills and executable action grounding through a hierarchical skill graph. Our data and code is available at https://github.com/BUPT-GAMMA/HiSkill.

Abstract

LLM Agent Skill Retrieval

Explore similar work

HiSkill: Empowering LLM Agents with Hierarchical Skill Graphs

HaReCAP: Habitual-action Grounding for Recursive Large Language Model Agents

SCALAR: Learning and Composing Skills through LLM Guided Symbolic Planning and Deep RL Grounding