PFIS-V: Modeling Foraging Behavior in the Presence of Variants
summary
A predictive model named PFIS-V extends PFIS3 by merging variant subgraphs into a single radiant space and modeling text-based equivalence and patch movement; it evaluates against ~500+ navigations from seven novice programmers using top‑10 hit rate. PFIS-V improves prediction accuracy — up to 25% more accurate than PFIS3 — and raises known‑patch coverage by treating knowledge transferred across variants, but evaluations reveal remaining errors because non‑lexical cues (visual output, runtime behavior) are not modeled.
AI-written summary · about
chapters (11)
- 0:00Intro
- 0:45The Problem
- 1:24Foraging Theory
- 2:22Foraging Variants
- 3:03Computational Models
- 4:43PFISV
- 5:32Data Models
- 8:37Navigation Behavior
- 9:45Models
- 10:39Best Model
- 13:15Summary
related talks
18mForaging among an Overabundance of Similar Variants
28mBPF Verifier Visualizer
21mSPINFER: Inferring Semantic Patches for the Linux Kernel
23mPrecise and Accurate Patch Presence Test for Binaries
11mIDFuzz: Intelligent Directed Grey-box Fuzzing
08m"How I use SFDIPOT to learn and explore about applications"
28mMPI: Multiple Perspective Attack Investigation...
20mFast Direct Manipulation Programming with Patch-Reconciliation Correspondence