MA-Bench: Towards Fine-grained Micro-Action Understanding

MA-Bench: Towards Fine-grained Micro-Action Understanding | Signal Canvas | ScienceToStartup

Page Freshness

Signal Canvas proof surface

Canonical route: /signal-canvas/ma-bench-towards-fine-grained-micro-action-understanding

stale

Proof freshness: stale
Proof status: unverified
Display score: 7/10
Last proof check: 2026-03-30
Score updated: 2026-04-02
Score fresh until: 2026-05-02
References: 96
Source count: 3
Coverage: 50%

This page is showing the last landed evidence receipt and score bundle because the latest proof data is outside the freshness window.

Agent Handoff

Canonical ID ma-bench-towards-fine-grained-micro-action-understanding | Route /signal-canvas/ma-bench-towards-fine-grained-micro-action-understanding

REST example

curl https://sciencetostartup.com/api/v1/agent-handoff/signal-canvas/ma-bench-towards-fine-grained-micro-action-understanding

MCP example

{
  "tool": "search_signal_canvas",
  "arguments": {
    "mode": "paper",
    "paper_ref": "ma-bench-towards-fine-grained-micro-action-understanding",
    "query_text": "Summarize MA-Bench: Towards Fine-grained Micro-Action Understanding"
  }
}

source_context

{
  "surface": "signal_canvas",
  "mode": "paper",
  "query": "MA-Bench: Towards Fine-grained Micro-Action Understanding",
  "normalized_query": "2603.26586",
  "route": "/signal-canvas/ma-bench-towards-fine-grained-micro-action-understanding",
  "paper_ref": "ma-bench-towards-fine-grained-micro-action-understanding",
  "topic_slug": null,
  "benchmark_ref": null,
  "dataset_ref": null
}

Evidence Receipt

Route status: building

Claims: 12

References: 96

Proof: Verification pending

Freshness state: computing

Source paper: MA-Bench: Towards Fine-grained Micro-Action Understanding

PDF: https://arxiv.org/pdf/2603.26586v1

Source count: 3

Coverage: 50%

Last proof check: 2026-03-30T21:51:53.348Z

Signal Canvas receipt window

Watch and verify: MA-Bench: Towards Fine-grained Micro-Action Understanding

/buildability/ma-bench-towards-fine-grained-micro-action-understanding

Watchwatch

Subject: MA-Bench: Towards Fine-grained Micro-Action Understanding

Verdict

Watch

Verdict is Watch because viability or proof quality is intermediate and should be re-evaluated before execution.

Time to first demo

Preparing verified analysis

GitHub Code Pulse

No public code linked for this paper yet.

Claim map

Strong 12Mixed 0Weak 0

Evidencepartial
we present MA-Bench, a benchmark comprising 1,000 videos and a three-tier evaluation architecture that progressively examines micro-action perception, relational comprehension, and interpretive reasoning. MA-Bench contains 12,000 structured question-answer pairs
Implicationpartial
The abstract explicitly states the creation and composition of MA-Bench.
Verificationpartial
partial
Evidencepartial
a three-tier evaluation architecture that progressively examines micro-action perception, relational comprehension, and interpretive reasoning.
Implicationpartial
The abstract clearly outlines the hierarchical structure of the MA-Bench evaluation.
Verificationpartial
partial
Evidencepartial
The results of 23 representative MLLMs reveal that there are significant challenges in capturing motion granularity and fine-grained body-part dynamics.
Implicationpartial
The abstract summarizes the findings from testing MLLMs on the benchmark.
Verificationpartial
partial
Evidencepartial
To address these challenges, we further construct MA-Bench-Train, a large-scale training corpus with 20.5K videos annotated with structured micro-action captions for fine-tuning MLLMs.
Implicationpartial
The abstract describes the purpose and content of MA-Bench-Train.
Verificationpartial
partial
Evidencepartial
The results of Qwen3-VL-8B fine-tuned on MA-Bench-Train show clear performance improvements across micro-action reasoning and explanation tasks.
Implicationpartial
The abstract provides a specific example of performance improvement after fine-tuning.
Verificationpartial
partial
Evidencepartial
Micro-motion tracker extracts motion descriptors (i.e., motion vectors and coordinates) for each body part.
Implicationpartial
The 'Parsed Sections' describe the technical steps involved in creating the benchmark.
Verificationpartial
partial
Evidencepartial
These videos have an average duration of 2.12 seconds, with a maximum length of 5.01 seconds.
Implicationpartial
The 'Parsed Sections' provide specific statistics about the video durations in MA-Bench.
Verificationpartial
partial
Evidencepartial
we present MA-Bench, a benchmark comprising 1,000 videos and a three-tier evaluation architecture that progressively examines micro-action perception, relational comprehension, and interpretive reasoning. MA-Bench contains 12,000 structured question-answer pairs
Implicationpartial
The abstract explicitly states the creation and composition of MA-Bench.
Verificationpartial
partial
Evidencepartial
a three-tier evaluation architecture that progressively examines micro-action perception, relational comprehension, and interpretive reasoning.
Implicationpartial
The abstract clearly outlines the three-tier evaluation structure of MA-Bench.
Verificationpartial
partial
Evidencepartial
To address these challenges, we further construct MA-Bench-Train, a large-scale training corpus with 20.5K videos annotated with structured micro-action captions for fine-tuning MLLMs.
Implicationpartial
The abstract explicitly states the purpose and size of MA-Bench-Train.
Verificationpartial
partial
Evidencepartial
The results of Qwen3-VL-8B fine-tuned on MA-Bench-Train show clear performance improvements across micro-action reasoning and explanation tasks.
Implicationpartial
The abstract reports specific performance improvements after fine-tuning a particular model on the proposed training corpus.
Verificationpartial
partial
Evidencepartial
The results of 23 representative MLLMs reveal that there are significant challenges in capturing motion granularity and fine-grained body-part dynamics.
Implicationpartial
The abstract summarizes the findings from evaluating existing MLLMs on MA-Bench, highlighting specific challenges.
Verificationpartial
partial

Author intelligence and commercialization panels stay hidden until the proof receipt is verified, cites at least 3 references, includes at least 2 sources, and clears 50% coverage. The paper narrative and citation surfaces remain public while verification is pending.

MA-Bench: Towards Fine-grained Micro-Action Understanding

Use Signal Canvas as the narrative proof surface

Use this Signal Canvas via API or MCP

Signal Canvas proof surface