00

ChatFuzz

Tool ☆ WATCH ⎘ SHARE

First seen 6/14/2026

Last seen 9/1/2026

Evidence 22 chunks

01

NEIGHBORHOOD

5 nodes · 9 edges

graph · ChatFuzz · depth=1

02

RELATIONSHIPS

21 connections

GenHuzz compares with → 100% 4e evidence

GenHuzz is benchmarked against ChatFuzz on coverage metrics.

GoldenFuzz compares with → 100% 3e evidence

GoldenFuzz is compared against ChatFuzz as a baseline hardware fuzzer.

TheHuzz compares with → 94% 2e evidence

Direct quantitative comparison of coverage speed and levels.

ISA Disassembler uses → 93% 2e evidence

A disassembler filters invalid generations and provides rewards during RL.

Coverage feedback uses → 95% 2e evidence

Coverage metrics are used to compute rewards and steer input generation.

Interdependent data/control-flow entangled instruction generation implements → 90% 2e evidence

The approach emphasizes producing entangled instruction sequences rather than random ones.

LLM-based Input Generator ← part of 82% 2e evidence

The architecture of ChatFuzz contains an LLM-based input generation module.

Beyond Random Inputs: A Novel ML-Based Hardware Fuzzing ← evaluates 92% 2e evidence

It measures condition coverage and time to coverage on RocketCore and BOOM.

GenHuzz: An Efficient Generative Hardware Fuzzer ← evaluates 90% 2e evidence

The paper evaluates ChatFuzz as a baseline for comparison with GenHuzz.

SearchSYS compares with → 85% 2e evidence

ChatFuzz is compared with SearchSYS in related work as a similar LLM-based fuzzing approach.

Beyond Random Inputs: A Novel ML-Based Hardware Fuzzing ← introduces 98% 1e evidence

Authors present ChatFuzz as a new ML-based hardware/processor fuzzer.

Differential Fuzz Testing uses → 92% 1e evidence

Compares ISA and RTL traces to find discrepancies.

FuSS: Coverage-Directed Hardware Fuzzing with Selective Symbolic Execution ← mentions 90% 1e evidence

The paper mentions ChatFuzz as a related hardware fuzzing approach.

LiFU compares with → 100% 1e evidence

LiFU is compared against ChatFuzz in the evaluation.

Reinforcement learning (RL) uses → 96% 1e evidence

RL optimizes the input generator based on coverage rewards.

Large Language Model uses → 90% 1e evidence

ChatFuzz uses an LLM to automatically generate seeds for fuzz testing.

Fuzzing uses → 90% 1e evidence

ChatFuzz is a fuzzing tool that leverages LLMs to improve fuzz testing.

Machine code sequence generation implements → 92% 1e evidence

Its LLM generates machine code sequences used as fuzzing inputs.

hardware fuzzing implements → 96% 1e evidence

ChatFuzz is positioned as a hardware fuzzer aimed at testing processors.

Large Language Models (LLMs) uses → 97% 1e evidence

LLMs are used to learn machine language and produce instruction sequences.

Hardware fuzzing implements → 85% 1e evidence

ChatFuzz injects reinforcement learning into hardware fuzzing.