Agentopia on a Consumer GPU: A Reduced-Scale Long-Horizon Port with an 8B Model

Luo Huan

cs.MAAug 25, 2026

Agentopia on a Consumer GPU: A Reduced-Scale Long-Horizon Port with an 8B Model

Authors: Luo Huan

Organizations: Shenzhen University Shenzhen, China

Abstract

Large language model (LLM)-based multi-agent social simulation has demonstrated compelling results, but Agentopia was evaluated with 100 agents over 10 simulated years using Qwen3.5-397B-A17B, leaving the behavior of reduced-scale deployments on consumer hardware unclear. In this paper, we implement and evaluate a reduced-scale Agentopia port on a single NVIDIA RTX 5070 Ti(12 GB VRAM) using Qwen3-8B-AWQ, a 4-bit quantized model. We introduce three structural adaptations for this setting: (1) system-managed layered memory compression, (2) four activity blocks per simulated day, and (3) explicit physical- and mental-health state variables. Across three independent stochastic runs, two runs completed 52 weeks and the third completed 50 weeks before reaching the context limit, totaling 154 system-weeks (770 agent-weeks). No agent died,and no threshold-based health warning was logged; activity records containing at least one NO_RESPONSE field occurred at rates of 10.15-10.29% across runs. A 52-week memory-off run tied L2/L3 artifact production to layered memory; a separate 10-week comparison associated four daily time blocks with 2.72 times more finalized records and lower lexical duplication, but a higher missing-field rate. These comparisons do not support causal behavioral claims. We release validated configurations, derived audits, analysis scripts, aggregate figure data, and our implementation changes in a public fork; raw runs and initial persona data are excluded because their redistribution provenance is not fully resolved.

Figures & tables

Figure 1: Architecture of the reduced-scale Agentopia port. A shared Qwen3-8B-AWQ endpoint served by vLLM executes Role-agent and God-model calls on a 12 GB consumer GPU. The weekly loop retains planning, contact and scheduling, activity, review, and settlement, while the M/T/H adaptations add system-managed layered memory, four activity blocks per simulated day, and explicit physical- and mental-health state. Solid arrows denote simulation flow; dashed arrows denote context or control paths.

Figure 2: Weekly physical- and mental-health trajectories across the three full-system runs. Lines show the weekly mean over the five active agents, and shaded bands span the agent-level minimum and maximum. Dotted horizontal lines mark the physical-health warning threshold of 10 and mental-health warning threshold of 20. Run 3 ends after week 50.

Agent	Run 1 PH/MH	Run 2 PH/MH	Run 3 PH/MH
Whitfield	100/93	95/70	96/71
Morales	100/97	100/79	100/89
Vale	93/92	96/77	100/96
Voss	100/79	100/77	100/99
Vieri	100/95	100/100	100/91

Table 1: Final-checkpoint physical-health (PH) and mental-health (MH) scores. Both range from 0 to 100. Runs 1–2 end at week 52 and Run 3 at week 50; each cell reports PH/MH.

Run	Total records	Flagged records	Rate
Run 1	4,412	448	10.15%
Run 2	4,436	451	10.17%
Run 3	4,236	436	10.29%
Pooled	13,084	1,335	10.20%

Table 2: Activity-record-level NO_RESPONSE incidence. Total records reports the number of finalized activity records, and Flagged records reports the number containing at least one NO_RESPONSE field; each record is counted at most once.

Condition	Weeks	Rec./A-W	NR	Exact	Near
M (T off)	10	6.36	1.57%	57.60%	16.40%
M+T	10	17.28	7.75%	10.06%	6.45%

Table 3: Exploratory 10-week time toggle (one run per condition; health disabled). Rec./A-W is finalized records per agent-week; NR is activity-record-level NO_RESPONSE . Exact/near are lexical duplicate rates in unflagged solo primary texts.

Run	Weeks	End condition	Activity records	NR rate	Min. PH	Min. MH	D/W	L2 entries/agent
Run 1	52	Target reached	4,412	10.15%	80	59	0/0	47
Run 2	52	Target reached	4,436	10.17%	80	53	0/0	47
Run 3	50	Context boundary	4,236	10.29%	80	60	0/0	45
Aggregate	154	–	13,084	10.20%	80	53	0/0	–

Table 4: Run-level summary. NR is the activity-record-level NO_RESPONSE rate, D/W denotes deaths/threshold-based health warnings, and health minima cover all completed weeks. L2 entries/agent is the final-checkpoint count per active agent. Aggregate weeks are system-weeks, and aggregate NR is pooled over records.

Configuration	Runs	Weeks completed	Status or termination	Observed outcome
Full (M+T+H)	3	52, 52, 50	Planned horizon completed (2); context boundary (1)	Zero deaths/warnings; L2/L3 artifacts present
$-$ Memory (T+H)	1	52	Planned 52-week horizon completed	L2/L3 artifacts absent by configuration
$-$ Health (M+T)	2	21, 38	Incomplete: week 21, resume-path slowdown; week 38, WSL I/O failure	No complete 52-week health comparison available
Time toggle (M vs. M+T)	1 each	10, 10	10-week comparison complete; 52-week M+H not run	$2.72\times$ records/A-W; lower reuse, higher NR; descriptive only
All off	1	4	Planned 4-week baseline completed	Longitudinal comparison unsupported; health warning/death mechanisms disabled

Table 5: Completion status and observed outcomes for the full-system runs, ablation conditions, and 4-week all-off baseline. M, T, and H denote layered memory, four-block daily time, and health.

Large Language Model-Based Social Simulation

Large language model-based social simulation studies how language-model agents represent people and interact to produce social behavior.

See all 261 papers

Three Roles, One Model: Role Orchestration at Inference Time to Close the Performance Gap Between Small and Large Agents
Apr 13, 2026S. Aaron McClendon, Jorge Gallego-Feliciano, Stavros Zervoudakis +1Tool-Using Language Agents LLM Agent Workflow Optimization
Aimpoint Digital Labs, Atlanta, GA, USA · New York University, New York, NY, USA
Poor Man's Agentic Modeling: Simulating Large LLM-Agent Societies on a Laptop
Jul 19, 2026Igor ItkinMulti-Agent LLM Systems Surrogate Modeling
Independent Researcher, Tel Aviv, Israel
Agentopia: Long-Term Life Simulation and Learning in Agent Societies
Jun 5, 2026Xintao Wang, Sirui Zheng, Hongqiu Wu +10Multi-Agent LLM Systems Human Behavior Simulation
Fudan University · Independent Researcher · Johns Hopkins University +1

Apr 13, 2026J/K move · Enter open · S save

Three Roles, One Model: Role Orchestration at Inference Time to Close the Performance Gap Between Small and Large Agents

S. Aaron McClendon, Jorge Gallego-Feliciano, Stavros Zervoudakis, Antonios Saravanos

Aimpoint Digital Labs, Atlanta, GA, USA · New York University, New York, NY, USA

Tool-Using Language Agents LLM Agent Workflow Optimization Inference-Time Scaling

Open paper AskPDF arXiv

Large language model (LLM) agents show promise on realistic tool-use tasks, but deploying capable agents on modest hardware remains challenging. We study whether inference-time scaffolding alone, without any additional training compute, can improve the performance of a small model in complex multi-step environments. Operating on a single 24GB GPU, we evaluate Qwen3-8B on the AppWorld benchmark under both full-precision and 4-bit quantized configurations. Without any intervention, the raw model achieves just 5.4% (FP16) and 3.0% (AWQ) task goal completion. Guided by a systematic failure mode analysis, we introduce a three-tier inference scaffolding pipeline that deploys the same frozen model in three distinct roles: (1) a summarization model that preserves critical artifacts (tokens, credentials, API responses) while compressing dialogue history; (2) the main agent model that reasons over the compressed context; and (3) an isolated correction model that reviews and revises the agent's code output without access to conversation history, breaking repetitive failure loops. Applied to the same unmodified model, this scaffolding yields 8.9% (FP16) and 5.9% (AWQ) task goal completion, roughly doubling performance in both settings, with particularly strong gains on difficulty-1 tasks (15.8% to 26.3% FP16; 5.3% to 14.0% AWQ). On full-precision inference, our scaffolded 8B model surpasses DeepSeek-Coder 33B Instruct (7.1%) from the original AppWorld evaluation, demonstrating that structured inference-time interventions can make small models competitive with systems 4 times their size. We formalize the approach as a scaffolded policy over a frozen base model, three invocations of the same weights with different conditioning, drawing connections to test-time compute scaling and action-space shaping in reinforcement learning.

Abstract

Large Language Model-Based Social Simulation

Explore similar work

Three Roles, One Model: Role Orchestration at Inference Time to Close the Performance Gap Between Small and Large Agents

Poor Man's Agentic Modeling: Simulating Large LLM-Agent Societies on a Laptop

Agentopia: Long-Term Life Simulation and Learning in Agent Societies