Organizations: Shanghai Jiao Tong University Shanghai, China · Ant Group Hangzhou, China · MoE Key Lab of Artificial Intelligence, AI Institute, Shanghai Jiao Tong University Shanghai, China
Large language model agents reuse execution experience to improve reliability on complex tool-based tasks. Systems such as Sentry call large models to generate failure diagnoses before retrieving advice, adding inference cost to experience reuse. We present Synapse-Control, an agent runtime whose lightweight control plane selects advice and determines when to send it to the task-executing model. Rules shortlist relevant advice, and a lightweight 1.7B classifier fine-tuned on teacher-labeled state-suggestion pairs scores those candidates instead of generating a diagnosis. Micro-contracts encode reusable interface checks, such as required arguments or valid options, without prescribing complete task programs. A Prompt Value Table records earlier guidance and feedback to limit repeated suggestions through cooldown and token-budget checks. Across 1,308 attempts per method in AppWorld and WebShop, Synapse-Control achieves 270 successes (29.8% more than Base), replaces 4,987 generative diagnosis calls, and reduces auxiliary control time relative to Sentry by 50.2% and 51.0%, respectively.
Figures & tables
Figure 1: Sentry’s recovery workflow. Solid arrows show the guidance path; the dashed arrow publishes verified lessons.
Figure 2: Recovery requires both economical control and effective guidance. (a) Sentry’s actor request time is normalized to 100% within each benchmark; stacked segments show additional control work. (b) In the AppWorld development probe (342 attempts per condition), a larger actor-output cap increases Base completions but does not improve Sentry completions.
Figure 3: Synapse-Control overview. Rules filter and prepare advice (§ 3.2 ), the classifier scores it (§ 3.1 ), and PVT decides when to send it (§ 3.3 ). Host verification and lesson updates remain active; hard repair is unchanged.
Figure 4: Aligned supervision: collect target-actor states, reconstruct available candidate groups, label the exact payload, and fit the small classifier. Scenario or product-group partitions separate training, calibration, and classifier testing.
Figure 5: Illustrative rendering of the implemented WebShop option-state asset. Both views use the same visible evidence; Contract adds the concrete check.
Figure 6: More completions with lower amortized execution cost. Each benchmark normalizes completions and total task time per completion to Base (100%). The cost numerator includes all registered attempts. Both panels use the same methods and axes.
Figure 7: Control time is halved in both benchmarks. Replacing generative diagnosis with the 1.7B sidecar reduces auxiliary control time by 50.2% in AppWorld and 51.0% in WebShop. Both panels show the same four roles per registered attempt; verification and lesson publication remain active.
System
Knowledge
Admission
Decision
State
Voyager [ 1 ]
Code skills
Skill retrieval
Actor reuse
Skill bank
ACE [ 7 ]
Playbook
Context update
Reflect + curate
Entry counters
Sentry [ 4 ]
Recovery lessons
Need checking
Generated diagnosis
Verified lessons
Synapse- Control
Micro-contracts
Visible evidence
Pair classifier
PVT + budget
Table 6: Selected knowledge and runtime interfaces. Columns distinguish the stored asset, admission signal, decision mechanism, and persistent state.
Appendix figures & tables5 assets
Supplementary material from the paper’s appendix.
Appendix
Setting
AppWorld
WebShop
Actor / classifier
Qwen3.5-9B / Qwen3-1.7B
Qwen3.5-9B / Qwen3-1.7B
Actor serving
BF16; non-thinking; TP=1
BF16; non-thinking; TP=1
Actor output / action cap
768 tokens / 60 actions
384 tokens / 60 actions
Actor temperature
0.7
0.7
Service context / active sequences
40,960 tokens / one
40,960 tokens / one
Actor memory utilization
0.65
0.65
Appendix
Table 7: Corresponding deployment and training settings. Each classifier is trained independently from the common initial checkpoint and frozen before its final evaluation.
Stage
AppWorld
WebShop
Environment facts
Signatures, parameters, and return conventions in API documentation
Visible actions, option values, and page state
Generic references
Authored trigger and checking principles
Five environment-reference families with checking conditions
State binding
Current call, observation values, and entity samples
Requested options, selected values, and purchase readiness
Online experience
Recovery lessons published after host verification
Recovery lessons published after host verification
Porting interface
Observation adapter, documentation catalog, and domain calibration
Page adapter, action constraints, and domain calibration
Appendix
Table 8: Knowledge sources and state binding in both benchmarks. Static construction, current observations, and online publication have separate interfaces.
Figure 8: Serialized object capacity on 1/2/4/6 real streams. Sharing immutable assets reduces total store capacity by 59.3% at six workers. This is a CPU object-layout replay.
Main-study records
AppWorld
WebShop
Registered attempts
1,008
300
Soft hints delivered
2,276
462
Attempts with budget records
993
219
Over-budget deliveries in these records
0
0
Appendix
Table 17: Corresponding delivery and budget records from the two main experiments. Over-budget counts refer to the recorded attempts listed in each column.
AppWorld
WebShop
Metric
Base-Llama
Synapse-Llama
Base-Llama
Synapse-Llama
Registered attempts
54
54
54
54
Jointly scored pairs
47
47
53
53
Successes (paired)
0/47
0/47
0/53
3/53
Mean benchmark score (paired)
16.18
16.18
3.09
14.34
Requests / mean-score point
197.0
211.4
999.0
198.6
Appendix
Table 18: Frozen-controller transfer: task outcomes and score-normalized resource use. Cost rows divide the fixed 54-attempt workload’s resource use by its paired mean benchmark score ( IC ). M and K denote millions and thousands of tokens; dashes indicate no guidance mechanism.
Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences · University of Chinese Academy of Sciences · Shenzhen University of Advanced Technology +1