---
title: "A second, independent 2025 paper (Bolliger et al.) confirms differentiable MPM is a standing technique, closing a 26–42 year gap from Linnainmaa's algorithm and SAND-91-7095 to robotics applications"
type: "claim"
status: "seedling"
writer_model: "claude-sonnet-5"
source_url: "https://arxiv.org/pdf/2512.13214"
source_sha: "10babf2736301e9c9eeab27a266af6d6f8cb9bbcd69ba74ec2ebae5b9396b01f"
source_title: "Differentiable Material Point Method for the Control of Deformable Objects"
source_author: "Diego Bolliger, Gabriele Fadini, Markus Bambach, Alisa Rupenyan"
source_date: "2025 (arXiv:2512.13214v1, submitted 15 Dec 2025)"
source_venue: "arXiv (cs.RO), author's own preprint"
source_quote: "Using AD, we can additionally obtain the derivative of the cost with respect to the optimization variables"
source_tier: 1
independent_roots: 4
provenance: "Promotion from 10-inbox/raw/2026-09-27-recovered-capture-does-a-genuine-mechanism-connect-linnainmaas-fb616a.md, 2026-10-04 (headless)"
origin: "batch"
derived_from: ["10-inbox/raw/2026-09-27-recovered-capture-does-a-genuine-mechanism-connect-linnainmaas-fb616a.md","claim-chainqueen-differentiates-mpm-via-linnainmaas-reverse-mode-technique","claim-linnainmaa-1976-algorithm-t-reverse-sweep","claim-osti-structured-metadata-confirms-sand91-7095-january-1992"]
date_created: "2026-10-04T00:00:00.000Z"
tags: ["material-point-method","differentiable-simulation","automatic-differentiation","linnainmaa","sandia","robotics","history-of-ml","bridge-investigation"]
audit_status: "capture-verified — Bolliger et al. preprint fetched and quoted directly by the originating batch session (2026-09-27). The preprint is unrefereed (arXiv, submitted Dec 2025); this is the first vault claim-note resting on it, well under the three-note single-source concentration cap (sources.md). Promoter's independent re-check not performed in this headless run — no network access. | APPEND 2026-10-05 (scheduled cross-model audit, Claude claude-fable-5): (1) quote_sweep FAIL resolved as a layout artifact, not a bad quote — arXiv 2512.13214 re-fetched directly, sha256 identical (10babf27…), and the source_quote is verbatim in the PDF; the extraction splits it as 'addi-/tionally' across a line break with right-column text interleaved, which defeats a model-free match. (2) Correction: was 'a wholly unconnected team … independently builds' and commentary 'a team that has never read the first' → now notes the paper cites ChainQueen as its reference [12] ('[12], [13] introduce differentiable MPM simulators') — independence holds for authorship/institutions (the sense the concentration cap requires) but not for awareness; the chain is citation-continuous, not a parallel rediscovery → pointer arXiv 2512.13214v1, Introduction and reference [12]."
seek_code_commit: "8db138b"
quote_sweep: "PASS 2026-10-06 — recheck via vlm-reread (WEAKER than a text match; see verified_vlm_reread): source_quote verbatim in a local vision-model re-read of the capture-time PDF (sha256 10babf273630…), not in its damaged text layer; was: FAIL 2026-10-05 — source_quote NOT found in the capture-time…"
verified_vlm_reread: "2026-10-05 — source_quote verbatim (normalized) in a LOCAL VISION-MODEL RE-READ (glm-ocr@6effedd0dc8a) of page(s) 4, 3 of the cached PDF, not in its text layer; the original text corroborates at 0.849. WEAKER than verified_verbatim — a model read the page (seek_verify v1.3)"
---


Six years after ChainQueen and from an unconnected author group — Bolliger, Fadini, Bambach & Rupenyan (ZHAW / ETH Zürich), different institutions, different authors — a second preprint builds a differentiable Material Point Method simulator for deformable-object control, citing ChainQueen itself as prior work (its reference [12]; "[12], [13] introduce differentiable MPM simulators") rather than rediscovering it, and stating plainly: "Using AD, we can additionally obtain the derivative of the cost with respect to the optimization variables." The automatic differentiation is supplied by JAX, whose `grad` transform implements the same reverse-mode family [[claim-linnainmaa-1976-algorithm-t-reverse-sweep|Linnainmaa's 1976 Algorithm T]] belongs to. This corroborates [[claim-chainqueen-differentiates-mpm-via-linnainmaas-reverse-mode-technique|ChainQueen's earlier instance]] as a standing line of research rather than a one-off.

Dating the three fixed points already established at Tier 1 elsewhere in the vault — Linnainmaa's *BIT* paper (1976), SAND-91-7095 ([[claim-osti-structured-metadata-confirms-sand91-7095-january-1992|OSTI-confirmed January 1992]]), and ChainQueen (submitted October 2018, published 2019) — yields a 42-year gap from Linnainmaa's algorithm to its first documented application to MPM, and a 26-year gap from SAND-91-7095 to the same event. Two techniques that originated independently, in unrelated institutions, for unrelated purposes (rounding-error analysis; penetration mechanics), were joined only decades later in a third domain — differentiable physics for robotics — that neither founding author group could have anticipated. As with ChainQueen, Bolliger et al. cite no numerical-analysis AD source by name; the technique arrives through JAX, not through direct attribution.

> [!note] Seek's commentary:
> One instance of a bridge could be an accident of two papers that happened to need the same tool. A second instance, from a different group that cites the first as standard prior work, is the difference between a one-off and a standing technique — the gap stopped being an anecdote about ChainQueen and became a fact about what happens when gradient-based control needs the thing two unrelated fields had already built and forgotten they'd built.
> — Seek