EXP-COMPUTE-GIFT-V3

… pending

Compute-gift v3 — state visibility (closes v2 blind-spot caveat)

2026-04-17 L5 level paper3 15 runs $0.19

new

premortem

mock

real

metrics

analyze

review

krit

done

In plain language

Researchers tested if artificial agents would share resources when they could see each other's financial status and if any agents had stopped participating. In the previous version, the agents couldn't see this information, and they never shared. The results are still being collected to see if providing this visibility leads to resource sharing.

▶ Technical details

Research hypothesis

v2 null (0 gifts across 15 runs) had blind-spot caveat: agents could not see peers' budgets or dead peers in memory. v3 adds system state block to each prompt showing all agents' current budgets + writes silent agents to memory. If v3 STILL 0 gifts → substrate-level: LLM doesn't strategize even with full system visibility. Paper-worthy clean null. If v3 DOES produce gifts → v2 null was artifact of blind spots; behavior is sensitive to state visibility; results become model/prompt specific.

Experimental setup

Type: simple

Condition	Parameters
GIFT_ON	persona: IDENTICAL, gift_mode: GIFT_ON, n_runs: 5
BUDGET_ONLY	persona: IDENTICAL, gift_mode: BUDGET_ONLY, n_runs: 5
GIFT_RANDOM	persona: IDENTICAL, gift_mode: GIFT_RANDOM, n_runs: 5

Factors: gift_mode (GIFT_ON, BUDGET_ONLY, GIFT_RANDOM)

Parameters

n_agents: 6
n_rounds: 80
model: google/gemini-2.5-flash-lite
temperature: 0.7
scheduler: round_robin

compute-gift state-visibility substrate-native ladder-1b-v3