{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2609.00103", "kind": "paper", "url": "https://arxiv.org/abs/2609.00103"}], "authors": ["Shmuel Berman", "Jia Deng"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2609.00103", "source_url": "https://arxiv.org/abs/2609.00103"}, "categories": ["General AI", "Multimodal"], "description": {"en": "Evaluates vision-language model memory along efficiency, compression, and calibration axes beyond raw accuracy."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2609.00103", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2609.00103", "title": "ECCBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Princeton University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2609.00103", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2609.00103", "source_url": "https://openalex.org/W7206186844"}, "key": "claire-radar:2609.00103", "modality": null, "name": "ECCBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "false", "origin_record_id": "bm_eccbench_68383aba", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "0db5567f7de62c98aed0445c799e46b6516a31dd34e78fd0cf3cdd15eca5cb11", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-name-audit-deferred", "reviewed_at": "2026-09-02T05:23:11.715633Z", "source_url": "https://arxiv.org/abs/2609.00103"}, "publisher": null, "released": "2026-08-31", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2609.00103", "source_url": "https://arxiv.org/abs/2609.00103"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2609-00103", "source": "claire_radar", "source_benchmark_id": "2609.00103", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Multimodal", "attention": {"asOf": "2026-09-02", "attemptedAt": "2026-09-02T05:12:51.621571Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2609.00103", "observedAt": "2026-08-31T04:56:28.279469Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-09-02T05:12:51.621571Z", "state": "unavailable"}}, "status": "unavailable"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Novel memory evaluation dimensions and comparative findings across architectures likely attract attention from both VLM and agent research communities.", "score": 65}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-02T05:23:11.715633Z", "inputHash": "af2f3f4943980cf00270a40a04379fae42a3f7d591d0738257761a72172e6e27", "model": "deepseek-v4-pro"}, "curation": {"decisionReason": "Formally named benchmark with clear evaluation protocol and scoring axes, despite missing public artifact links in supplied evidence.", "model": "deepseek-v4-pro", "policy": "source-evidence-v2", "reviewedAt": "2026-09-02T05:23:11.715633Z", "state": "ai-name-audit-deferred"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates vision-language model memory along efficiency, compression, and calibration axes beyond raw accuracy.", "displayEligible": false, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce ECCBench, a benchmark and evaluation protocol that measures memory beyond a system's capacity--its raw accuracy at a specific budget--via three axes we call ECC: efficiency--the computation, in FLOPs, needed to answer from memory; compression--whether compressible inputs are remembered more accurately or efficiently; and calibration--whether the system abstains in response to its own uncertainty and the cost of an error."}, "familyId": "bmf_f9586e5ebcd8", "firstSeenAt": "2026-09-02", "heat": null, "id": "bm_eccbench_68383aba", "indexedAt": "2026-09-02T05:21:06.631990Z", "industrySectors": [], "links": {"code": null, "data": null, "paper": "https://arxiv.org/abs/2609.00103", "pdf": "https://arxiv.org/pdf/2609.00103", "project": null, "report": "https://arxiv.org/abs/2609.00103"}, "metrics": [], "motivation": "Memory is widely viewed as an important unsolved problem for LLMs and VLMs, and current benchmarks typically evaluate it by testing accuracy over long text or video.", "name": "ECCBench", "oneLine": "Evaluates vision-language model memory along efficiency, compression, and calibration axes beyond raw accuracy.", "paperTitle": "Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2609.00103", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-09-02T05:21:06.631990Z"}, "publications": [], "ranking": {"30d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}, "90d": {"confidence": "Low", "coverage": 0.0, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": null, "value": null}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "rank": null, "score": null}, "method": "attention-ranking-v13", "rank": null, "score": null}}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-31", "source": {"authors": ["Shmuel Berman", "Jia Deng"], "categories": ["cs.LG", "cs.AI"], "id": "2609.00103", "title": "Good Memory Has ECC: Evaluating the Memory of Vision-Language Models Beyond Accuracy", "type": "arxiv", "url": "https://arxiv.org/abs/2609.00103"}, "sourceUpdatedAt": "2026-08-31T17:05:17Z", "topics": ["Multimodal"], "venueAttempts": [], "whyItMatters": "Identifies memory limitations in VLMs and highlights non-Transformer architectures with better tradeoffs for long-horizon tasks."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
