Md. Asif Uddin

    Chapter 7 V.7

    VLM Evaluation

    Caption and answer metrics reward surface agreement, so a wrong answer can outscore a right one.

    Metrics reward surface agreement, and can be satisfied by a wrong answer.

    How this chapter is built

    M2Substantive

    The derivations are the chapter. A reader who skips the algebra has not learned it.

    basics2/9what the words mean
    concept2/2what to picture
    theory0/2why it works, and when it does not
    mathematics0/9derive it, then compute it
    practice0/6build it, break it, read the papers

    Five strands, not one. Mathematics is the spine; the other four are the body. A chapter cannot pay its way out of teaching with problems, nor out of problems with teaching.

    Before you start

    The problem

    Generation has no single correct output, so every metric is a proxy. Knowing which proxy rewards what — and constructing the case where it rewards the wrong thing — is the only defence against a leaderboard.

    Two evaluations, opposite conclusionsArc reports that STATE is the first model in its domain to consistently beat simple linear baselines. VCBench reports that pre-registered baselines match or exceed every foundation model tested on four of five dimensions. Four ordinary differences account for the gap.Arc, June 2025first to consistently beatsimple linear baselinesVCBench, June 2026baselines match or exceed allfive models on four of fivebothin printdifferent baselinesdifferent splitsdifferent metricsa year apartNeither claim is dishonest. Working out how they can both be true is more useful than picking a side.If a benchmark cannot tell interaction structure from main effects, a good score on it demonstrates little.
    Fig. 7 — Two credible evaluations pointing opposite ways, and the four ordinary differences that account for it. Neither claim is dishonest.

    What this chapter covers

    • VQA
    • Captioning
    • Grounding
    • Hallucination
    • Multimodal reasoning
    • Robustness

    Apparatus

    The mathematics this chapter leans on, held in Book 0 so it can be assumed here without being taught here. Not a gate — follow a link when a step stops making sense.

    Hypothesis tests 0.ST.03 · Confidence intervals 0.ST.02

    Notation

    • 𝔼Expectation
    • θ̂An estimate, as against the quantity it estimates

    Propositions

    Not yet written. The topics above are the plan for this chapter; each will become a proposition with its own figure.

    Worked problems

    0/3 problems0/3 variants0/6 exercisesowes 9 more

    Not yet written. At M2 this chapter owes 3 worked problems across 3 distinct variants, and 6 exercises, every one with a published solution. The build enforces that from the day the chapter is marked published.