{ "schema_version": 1, "builder_plan": { "scenario": "A self-hosted model uses more compute or memory than your workload can justify.", "change": "Check for a usable pruned checkpoint and runtime support before attempting model surgery.", "baseline": "The original unpruned model at the same precision, context, and load.", "steps": [ "Locate the paper's implementation and check its license, model support, data needs, and hardware requirements. Stop here if you cannot obtain the required artifacts.", "Save a small set of representative product tasks and expected outcomes. Keep evaluation tasks out of any training or tuning, and choose quality and cost limits before running.", "Run the baseline and the proposed change on identical inputs. Keep other settings fixed and repeat paired runs to expose variation.", "Save per-task outputs, failures, timings, and all usage costs. Review regressions before expanding the test." ], "measure": "Task quality, actual GPU memory, serving throughput, and preparation cost.", "decision": "Keep the change only if it meets your preselected quality and latency limits and improves the metric you care about after setup and operating costs. Otherwise retain the baseline. This is an exploration plan, not a validated recommendation." }, "builder_review": null, "arxiv_id": "2610.03389", "title": "From Patching to Pruning Visual Computation in Vision Language Models", "summary": "P2P replaces token-specific visual projection outputs in selected decoder layers with fixed neutral proxy activations, identified by validation-guided forward and backward layer sweeps, rather than pruning tokens or altering weights, so sequence length, positions, and attention masks remain intact. It costs a one-time calibration procedure on disjoint validation partitions, an accepted accuracy tolerance (3% in the paper, retaining about 94% of dense accuracy), and added per-layer routing logic in the inference path. It can fail when the proxy substitution tolerance is mis-set for a new task or domain, when calibration data is unrepresentative of production inputs, or on models and benchmarks outside the evaluated Qwen2.5-VL and LLaVA families where the layer-sensitivity pattern may not hold.", "rationale": "The method is training-free and runs on a single GPU with the reader's likely VLM stack, but the accuracy-efficiency trade-off depends on per-workload calibration and tolerance selection, so it should be validated on the reader's own tasks before production adoption.", "technique": "Layer-wise visual computation bypass via neutral proxy activations", "family": "esparsidade_e_poda", "recommendation": "testar", "published": "2026-10-02", "as_of": "2026-10-18", "editorial_status": "indexed", "source_url": "https://arxiv.org/abs/2610.03389", "independent_implementations": 0, "total_implementations": 1, "stars": 121, "citations": null, "claims": [ { "claim_id": "claim-01", "statement": "P2P replaces token-specific visual projection outputs in selected decoder layers with fixed neutral proxy activations, identified by validation-guided forward and backward layer sweeps, rather than pruning tokens or altering weights, so sequence length, positions, and attention masks remain intact. It costs a one-time calibration procedure on disjoint validation partitions, an accepted accuracy tolerance (3% in the paper, retaining about 94% of dense accuracy), and added per-layer routing logic in the inference path. It can fail when the proxy substitution tolerance is mis-set for a new task or domain, when calibration data is unrepresentative of production inputs, or on models and benchmarks outside the evaluated Qwen2.5-VL and LLaVA families where the layer-sensitivity pattern may not hold.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.03389", "source_page": null, "source_excerpt": "" }, { "claim_id": "claim-02", "statement": "Reduces inference FLOPs by 55% while retaining approximately 94% of dense accuracy at a 3% tolerance across seven multimodal benchmarks on Qwen2.5-VL and LLaVA models.", "basis": "inferred", "result": "", "baseline": "", "conditions": "", "source_url": "https://arxiv.org/abs/2610.03389", "source_page": null, "source_excerpt": "" } ], "exposure_map": [ { "dimension": "quality", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compute", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "latency", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "operations", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "compatibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "security", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "data_and_training", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" }, { "dimension": "reproducibility", "basis": "not_evaluated", "finding": "", "interpretation": "", "limitation": "", "next_check": "", "source_url": "", "source_page": null, "source_excerpt": "" } ], "risks": [], "minimum_test": [], "open_questions": [], "independent_tests": [], "lab_reproductions": [], "report_available": false, "validation_tier": "unknown", "evidence_tier": "unknown", "infrastructure_basis": "unknown", "equations": [], "equations_status": "not_fetched", "core_kind": "", "equations_fetched_at": "", "equations_source_url": "", "equations_source_sha256": "", "exposures_reviewed_at": "", "exposures_source_sha256": "" }