Pith Number

pith:Z4QTKRP5

pith:2026:Z4QTKRP57TX45TFZGCOKCUYWGV

not attested not anchored not stored refs resolved

Hide to See: Reasoning-prefix Masking for Visual-anchored Thinking in VLM Distillation

Byung-Kwan Lee, Dongjun Nam, Jeany Son, Seonghoon Yu

Masking the student's salient reasoning prefixes during distillation makes VLMs anchor their thinking more directly on visual evidence.

arxiv:2605.11651 v3 · 2026-05-12 · cs.CV · cs.AI · cs.CL

Open paper page JSON Open Graph Bundle Merged state Verified badge What is a Pith Number?

Add to your LaTeX paper

\usepackage{pith}
\pithnumber{Z4QTKRP57TX45TFZGCOKCUYWGV}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp

2 Internet Archive

3 Author claim open · sign in to claim

4 Citations open

5 Replications open

✓ Portable graph bundle live · download bundle · merged state

The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

Our approach outperforms recent open-source VLMs, VLM distillation, and self-distillation methods on multimodal reasoning benchmarks, while further analyzes confirm enhanced visual utilization along the student thinking process.

C2weakest assumption

That masking the student's salient reasoning prefixes will cause the student to rely more on visual evidence as an alternative source of information during distillation, as stated in the abstract's description of the framework objective.

C3one line summary

A masking-based think-answer distillation method for VLMs that selectively hides reasoning prefixes and uses self-paced scheduling to improve visual anchoring and benchmark performance.

References

40 extracted · 40 resolved · 9 Pith anchors

[1] On-policy distillation of language models: Learning from self-generated mistakes 2024

[2] Qwen3-VL Technical Report 2025 · arXiv:2511.21631

[3] Qwen2.5-VL Technical Report 2025 · arXiv:2502.13923

[4] Llava-kd: A framework of distilling multimodal large language models 2025

[5] Move-kd: Knowledge distillation for vlms with mixture of visual encoders 2025

Formal links

2 machine-checked theorem links

Receipt and verification

First computed	2026-05-20T00:00:42.871120Z
Builder	pith-number-builder-2026-05-17-v1
Signature	Pith Ed25519 (`pith-v1-2026-05`) · public key
Schema	pith-number/v1.0

Canonical hash

cf213545fdfcefceccb9309ca15316356d8aa20be189b8a0c89391981e088970

Aliases

arxiv: 2605.11651 · arxiv_version: 2605.11651v3 · doi: 10.48550/arxiv.2605.11651 · pith_short_12: Z4QTKRP57TX4 · pith_short_16: Z4QTKRP57TX45TFZ · pith_short_8: Z4QTKRP5

Agent API

Resolver JSON Graph JSON Events JSON Schema Signing key

Verify this Pith Number yourself

curl -sH 'Accept: application/ld+json' https://pith.science/pith/Z4QTKRP57TX45TFZGCOKCUYWGV \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: cf213545fdfcefceccb9309ca15316356d8aa20be189b8a0c89391981e088970

Canonical record JSON

{
  "metadata": {
    "abstract_canon_sha256": "7ef022a8ae382b90ca80b056b34399afef3aeb983bf4cdd7a3b24eeadebcab68",
    "cross_cats_sorted": [
      "cs.AI",
      "cs.CL"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.CV",
    "submitted_at": "2026-05-12T07:14:04Z",
    "title_canon_sha256": "b6230ede9c38f08c7567fc71f0fd309a8b88e37af49e812bb2d03d94a9a36d99"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.11651",
    "kind": "arxiv",
    "version": 3
  }
}