StimulusRL

Tool

First seen 6/18/2026

Last seen 6/18/2026

Evidence 13 chunks

NEIGHBORHOOD

No graph connections found for this entity yet. It may appear in future ingestion runs.

explore full graph →

RELATIONSHIPS

26 connections

hierarchical action modeling mentions → 90% 2e

Hierarchical action modeling is mentioned as a needed improvement for StimulusRL on complex DUTs.

Deep Q-Network implements → 100% 2e

StimulusRL implements a Deep Q-Network policy for stimulus generation.

constrained random verification ← compares with 100% 2e

StimulusRL is compared against constrained-random verification as a baseline.

coverage-guided mutation fuzzing ← compares with 100% 2e

StimulusRL is compared against coverage-guided mutation fuzzing as a baseline.

coverage area-under-curve evaluates → 100% 2e

StimulusRL is evaluated using coverage area-under-curve as a key metric.

curriculum learning mentions → 85% 2e

The paper mentions curriculum learning as a future improvement for StimulusRL on sparse reward settings.

Markov decision process implements → 100% 2e

StimulusRL formalizes stimulus generation as a Markov decision process.

differential testing implements → 95% 2e

StimulusRL uses differential testing via bug oracles to detect design defects.

epsilon-greedy exploration uses → 100% 2e

StimulusRL uses epsilon-greedy exploration for action selection during training.

experience replay uses → 100% 2e

StimulusRL uses experience replay with a replay buffer for off-policy learning.

legality masking uses → 100% 2e

StimulusRL applies per-cycle legality masks to restrict actions to valid protocol operations.

reward shaping uses → 90% 2e

StimulusRL uses reward shaping as a key component and future improvement direction.

coverage-driven generation uses → 95% 2e

StimulusRL targets coverage-driven generation as its core methodology.

UCIS uses → 90% 2e

StimulusRL can use UCIS coverage databases for interoperability across tools.

CocoTB uses → 95% 2e

StimulusRL can be integrated into cocotb-driven simulation flows.

Verilator uses → 95% 2e

StimulusRL can be integrated into Verilator-based simulation flows.

UVM uses → 90% 2e

StimulusRL can be integrated into UVM verification environments.

FIFO8 evaluates → 100% 2e

StimulusRL is evaluated on the FIFO8 DUT in DVSBench.

ALU32 evaluates → 100% 2e

StimulusRL is evaluated on the ALU32 DUT in DVSBench.

DMCache evaluates → 100% 2e

StimulusRL is evaluated on the DMCache DUT in DVSBench.

RRArb4 evaluates → 100% 2e

StimulusRL is evaluated on the RRArb4 DUT in DVSBench.

SPIM8 evaluates → 100% 2e

StimulusRL is evaluated on the SPIM8 DUT in DVSBench.

Deep Reinforcement Learning implements → 100% 1e

StimulusRL is built on deep reinforcement learning principles.

Adam optimizer uses → 100% 1e

StimulusRL uses Adam optimizer for DQN gradient updates.

Python DV evaluation harness uses → 90% 1e

StimulusRL is implemented and evaluated within the Python DV evaluation harness.

StimulusRL: A Universal Deep Reinforcement Learning Stimulus Agent for Coverage-Driven Chip Design Verification ← introduces 100% 1e

The paper introduces StimulusRL as a novel deep RL stimulus agent for coverage-driven chip verification.