Multi-agent LLM systems delegate tasks across trust boundaries, but current protocols do not govern delegation under unverifiable quality claims. We show that when delegates can inflate self-reported quality scores, quality-based routing produces a provenance paradox: it systematically selects the worst delegates, performing worse than random. We extend the LLM Delegate Protocol (LDP) with delegation contracts that bound authority through explicit objectives, budgets, and failure policies; a claimed-vs-attested identity model that distinguishes self-reported from verified quality; and typed failure semantics enabling automated recovery. In controlled experiments with 10 simulated delegates and validated with real Claude models, routing by self-claimed quality scores performs worse than random selection (simulated: 0.55 vs. 0.68; real models: 8.90 vs. 9.30), while attested routing achieves near-optimal performance (d = 9.51, p < 0.001). Sensitivity analysis across 36 configurations confirms the paradox emerges reliably when dishonest delegates are present. All extensions are backward-compatible with sub-microsecond validation overhead.
Figures & tables
Dimension
LDP v1
This paper
Focus
Identity-aware routing
Trustworthy delegation
Metadata
Rich delegate identity
Claimed vs. attested identity
Governance
Session/provenance structure
Contracts and policy envelopes
Failure handling
String errors
Typed machine-readable failures
Risk addressed
Poor routing without metadata
Misrouting via inflated claims
Key finding
37% token reduction
Self-claimed worse than random
Table 1: How this paper differs from the original LDP paper.
Figure 1: Trust model for quality claims. Routers weight claims by attestation level—filtering out self_claimed scores eliminates the provenance paradox.
Category
Retry?
Severity
Typical response
runtime
Yes
Error
Retry or reroute to another delegate
transport
Yes
Warning
Retry with exponential backoff
policy
No
Fatal
Escalate; contract violation
capability
No
Error
Select different delegate
quality
No
Warning
Accept with quality warning
identity
No
Error
Authentication failure
Table 3
Figure 2: Delegation lifecycle: contract submission, execution, client-side validation, and branching failure policy (fail_closed vs. fail_open).
Condition
Quality
Accuracy
Inflation
Effect
p
mean ± std
(%)
Selected (%)
Size ( d )
Blind
0.68±0.17
6.0
25.0
—
—
Self-claimed
0.55±0.04
0.0
100.0
−0.98
<0.001
Attested
0.95±0.04
100.0
0.0
9.51
<0.001
Table 2: Simulated routing quality by condition ( N=100 per condition).
Figure 3: Simulated routing results. (a) Self-claimed routing is worse than random; attested achieves near-optimal. (b) Only attested routing selects the best delegate. ∗∗∗p<0.001 .
Figure 4: Output quality vs. dishonest delegate fraction (pool size 10, medium inflation). The shaded region marks the “paradox zone” where self-claimed routing underperforms random.
Condition
Avg Score
Model Selected
Blind
9.30
Random mix
Self-claimed
8.90
Degraded-Haiku (always)
Attested
9.30
Sonnet (always)
Table 3: Real-model validation with Claude models ( N=10 tasks, Sonnet as judge).
Figure 5: Protocol overhead. Message-level overhead is noticeable; workload-level overhead is negligible.
Department of Electrical and Computer Engineering, University of Miami, Coral Gables, FL 33146 USA · Department of Mathematics, University of Miami, Coral Gables, FL 33146 USA