LiFU

Tool

First seen 8/5/2026

Last seen 8/5/2026

Evidence 15 chunks

NEIGHBORHOOD

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

explore full graph →

RELATIONSHIPS

50 connections

ProcessorFuzz compares with → 100% 2e

LiFU is compared against ProcessorFuzz in the evaluation.

RTL uses → 100% 2e

LiFU ingests RTL design hierarchy and runs RTL simulation.

Weight Cache Seed Scoring implements → 100% 2e

LiFU uses a Weight Cache to dynamically score and prioritise seeds.

Dependency Chain Generation implements → 90% 2e

LiFU generates instruction sequences with dependency chains to trigger micro-architectural hazards.

Assembly Synthesis implements → 100% 2e

LiFU synthesises assembly instruction sequences using LLMs.

Human-in-the-Loop Triage uses → 90% 2e

LiFU involves human experts in mismatch triage and annotation.

mismatch detection uses → 100% 2e

LiFU detects mismatches between DUT and ISS outputs.

Non-Progressing Testcase Filtering uses → 100% 2e

LiFU filters out non-progressing testcases to avoid wasted RTL simulation cycles.

Instruction Diversity uses → 90% 2e

LiFU aims to achieve high instruction diversity to cover the micro-architectural state space.

Poisonous Interaction uses → 100% 2e

LiFU detects and avoids poisonous interactions between fuzzer combinations.

FSM coverage uses → 100% 2e

LiFU measures and improves FSM coverage on CPU designs.

line coverage uses → 100% 2e

LiFU measures and improves line coverage on CPU designs.

condition coverage uses → 100% 2e

LiFU measures and improves condition coverage on CPU designs.

Formal Verification implements → 100% 2e

LiFU integrates formal verification via VC Formal for property proving.

Feedback-Driven Coordination implements → 100% 2e

LiFU uses feedback-driven coordination to steer fuzzers in real time.

SVA Property Synthesis implements → 100% 2e

LiFU synthesises SVA properties via LLM for uncovered coverpoints.

Pipeline Hazard uses → 90% 2e

LiFU is designed to trigger pipeline hazards through directed mutation.

spike uses → 100% 2e

LiFU uses Spike ISS as a golden oracle for differential testing.

ISS Pre-Run Filtering implements → 100% 2e

LiFU filters non-progressing testcases via ISS pre-runs.

Reachability Analysis implements → 100% 2e

LiFU performs reachability analysis on uncovered RTL coverpoints.

Cascade uses → 100% 2e

LiFU adopts Cascade for seed mutation as part of its ensemble.

LLM-Guided Property Generation implements → 100% 2e

LiFU generates SVA properties for hard-to-reach coverpoints using LLMs.

Large Language Model (LLM) uses → 100% 2e

LiFU leverages LLMs for generative mutation and property generation.

Knowledge Base uses → 95% 2e

LiFU maintains and queries a knowledge base to ground LLM prompts.

Coverage Feedback uses → 100% 2e

LiFU uses coverage feedback to steer mutation and seed selection.

seed corpus uses → 100% 2e

LiFU maintains and evolves a seed corpus throughout the fuzzing campaign.

RocketChip evaluates → 100% 2e

LiFU is evaluated on the RocketChip RISC-V processor core.

BOOM evaluates → 100% 2e

LiFU is evaluated on the BOOM out-of-order RISC-V core.

differential testing implements → 100% 2e

LiFU performs differential checking against ISS golden traces.

Pipelined Fuzzing implements → 100% 2e

LiFU adopts a pipelined design to maximise testcase throughput.

Ensemble Fuzzing implements → 100% 2e

LiFU orchestrates multiple fuzzers as an ensemble.

Cascade ← compares with 100% 2e

LiFU is compared against Cascade in the evaluation.

Data Forwarding uses → 85% 1e

LiFU is designed to expose data forwarding hazards in processor pipelines.

Smart Coordination implements → 100% 1e

LiFU implements smart coordination among multiple fuzzers.

Semantic Seed Triage implements → 95% 1e

LiFU uses semantic seed triage to augment its coordination.

LLM-Guided Generative Mutation implements → 100% 1e

LiFU uses LLM-powered generative mutation targeting micro-architectural gaps.

Asymmetric Simulation implements → 100% 1e

LiFU adopts an asymmetric simulation strategy decoupling ISS from RTL execution.

Coverage-Guided Fuzzing implements → 95% 1e

LiFU uses coverage feedback to guide fuzzing.

Multi-Armed Bandit implements → 90% 1e

LiFU adapts seed scoring coefficients via multi-armed bandit online feedback.

VC Formal uses → 100% 1e

LiFU uses VC Formal for formal verification of LLM-generated properties.

GPT-4-turbo uses → 100% 1e

LiFU uses GPT-4-turbo via OpenAI API for assembly synthesis.

ProcessorFuzz uses → 100% 1e

LiFU adopts ProcessorFuzz for seed mutation as part of its ensemble.

DiFuzzRTL ← compares with 100% 1e

LiFU is compared against DifuzzRTL in the evaluation.

ChatFuzz ← compares with 100% 1e

LiFU is compared against ChatFuzz in the evaluation.

GenHuzz ← compares with 100% 1e

LiFU is compared against GenHuzz in the evaluation.

RISC-V uses → 100% 1e

LiFU is applied to RISC-V processor cores.

Algorithm 1: Filter for Non-Progressing Testcases ← part of 100% 1e

LiFU includes Algorithm 1 for filtering non-progressing testcases.

Algorithm 2: Reachability Analysis for Uncovered Point ← part of 100% 1e

LiFU includes Algorithm 2 for reachability analysis of uncovered coverpoints.

Algorithm 3: LLM-Guided Property Generation ← part of 100% 1e

LiFU includes Algorithm 3 for LLM-guided property generation.

CPU fuzzing implements → 100% 1e

LiFU is a CPU fuzzing framework.