2026 · paper
A source-free segmentation quality estimation of a model adapted to a new domain
Studies quality estimation for an adapted segmentation model when source data are unavailable.[75]
Statistical machine learning and multimodal inference
Jacob Goldberger
יעקב גולדברגר
Identity: verifiedThe BIU faculty page and linked CV use Jacob Goldberger and align with the roster. [75][76]
Documented foundation
Research & experience
Goldberger develops statistical and deep-learning methods across vision, audio, medical imaging, and language.[75][76]
deep learningmedical image analysisspeech and audio modelingnatural-language processinguncertainty estimation
CV and official profile
A current BIU-hosted CV was opened and linked; only bounded career and output facts were used.[76]
Open CVSelected work
Representative records, not a complete publication list. Metadata confirms attribution; it does not independently replicate a result.
Patent evidence
11 catalogued patent records · 10 identified families · family unassigned for 1 record
Coverage: Partial inventor search
Ten additional family representatives span ART, CogniTens, CUTe, Varonis, and academic/medical assignees. CV corroborates older employers and six patent numbers; CUTe applications are supported by matching employment period and technical area. Existing mammogram patent retained. Similar titles were not merged unless an inspected family ID matched.
Permutation selection for decoding of error correction codes
US20220231785A1 · Published 2022-07-21
Published patent document inspected
Publication assignee: Ramot at Tel Aviv University Ltd; Bar Ilan University
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][257]
Method and system of classifying medical images
US9122955B2 · Published 2015-09-01
Published patent document inspected
Publication assignee: Ramot at Tel Aviv University Ltd; Bar Ilan University; Tel HaShomer Medical Research Infrastructure and Services Ltd
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][258]
Automatic management of storage access control
US7606801B2 · Published 2009-10-20
Published patent document inspected
Publication assignee: Varonis Inc
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][259]
System and method for feedback-based unequal error protection coding
US20020157058A1 · Published 2002-10-24
Published patent document inspected
Publication assignee: CUTe Ltd
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][264]
Unequal error protection of variable-length data packets based on recursive systematic convolutional coding
US20020146074A1 · Published 2002-10-10
Published patent document inspected
Publication assignee: CUTe Ltd
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][265]
Show 6 moreShow fewer patent records
System for enhanced error correction in trellis decoding
US20020144209A1 · Published 2002-10-03
Published patent document inspected
Publication assignee: CUTe Ltd
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][266]
System and method for “Stitching” a plurality of reconstructions of three-dimensional surface features of object(s) in a scene defined relative to respective coordinate systems to relate them to a common coordinate system
US6201541B1 · Published 2001-03-13
Published patent document inspected
Publication assignee: Cognitens Ltd
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][261]
Pattern recognition system
US6195638B1 · Published 2001-02-27
Published patent document inspected
Publication assignee: ART Advanced Recognition Technologies Inc
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][260]
Handwritten pattern recognizer with selective feature weighting
US6023529A · Published 2000-02-08
Published patent document inspected
Publication assignee: ART Advanced Recognition Technologies Inc
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][262]
Pattern recognition system
US5809465A · Published 1998-09-15
Published patent document inspected
Publication assignee: ART Advanced Recognition Technologies Inc
Named inventor Jacob Goldberger; identity corroborated by the CV's patent entries or CUTe employment and matching co-inventors/technical area. Assignee is the captured publication metadata, not a current-ownership determination.[237][263]
Weakly and fully labeled mammogram classification and localization with a dual branch deep neural network
US10789462B2
Original report record
Preserved from the original report; see its cited evidence and limitations.[77]
Original report snapshot
Original evidence: verified record
The patent record verifies inventor attribution; present legal status or ownership is not used to infer research freedom.[77]
Records are counted separately from identified families. Author-reported entries are labelled and may still need publication verification. Inventorship, publication-time applicant and current ownership are different facts. No legal-status, patentability or freedom-to-operate conclusion is made.
Scores prioritize research fit from 1–10; they are not probabilities.
Review: Reviewed with limitations
Proposed capability matches, not confirmed relationships. Scores are analyst judgments with low forecast confidence; researcher interests, capacity and feasibility need confirmation.
Internal connections
15 candidates
Connection 1
Yaara Erez
Original proposal
Proposal hypothesis: Erez's electrophysiology and brain-network analyses provide neural targets for Goldberger's statistical learning and uncertainty methods.[28][29][75][76]
First test and score details
First test
Proposed first test: Evaluate one public neural decoding task with held-out subjects; compare calibrated abstention and accuracy with a linear signal-analysis baseline.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 1/15; fit 10/10 (4 topic overlap + 3 complementarity + 3 feasible first test). Preserved original co-membership proposal in o05. No strictly higher-scoring candidate displaces this original. Strong data-method fit and public benchmark feasibility; decoding does not establish causal brain function. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Original initiative conditions remain: o05 Jacob Goldberger: Conditional later role: evaluate uncertainty after a reference dataset and transparent signal-analysis baseline exist.
Connection 2
Ethan Fetaya
Original proposal
Proposal hypothesis: Fetaya's adversarial and Bayesian learning and Goldberger's domain-shift and uncertainty work meet on detecting confidently wrong predictions.[73][75][76]
First test and score details
First test
Proposed first test: Apply a fixed corruption suite to one public image benchmark and compare calibration, abstention and retained accuracy against an uncalibrated baseline.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 2/15; fit 10/10 (4 topic overlap + 3 complementarity + 3 feasible first test). Preserved original co-membership proposal in o01. No strictly higher-scoring candidate displaces this original. Direct methodological overlap and a bounded benchmark justify a high test-fit score, without claiming robustness in deployment. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 3
Shahar Alon
Proposal hypothesis: Alon's spatial sequencing images and Goldberger's medical-image uncertainty methods meet on detecting unreliable molecular image analysis.[60][61][75][76]
First test and score details
First test
Proposed first test: Apply controlled image degradation to an open spatial-transcriptomics image set and compare segmentation quality flags with a simple image-quality baseline.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 3/15; fit 10/10 (4 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Molecular localization ground truth and platform-specific image artifacts need explicit definitions. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Show 12 moreShow fewer internal connections
Connection 4
Tomer Kalisky
Proposal hypothesis: Kalisky's single-cell state questions and Goldberger's statistical learning support uncertainty-aware cell-state classification.[70][75][76]
First test and score details
First test
Proposed first test: Hold out one public single-cell study and compare calibrated and uncalibrated state assignments on rare-cell recall and rejection rate.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 4/15; fit 10/10 (4 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. No clinical interpretation or mechanistic inference follows from classification accuracy alone. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 5
Nisan Ozana
Original proposal
Proposal hypothesis: Ozana's diffuse optical measurements have signal-quality and domain-shift questions suited to Goldberger's uncertainty and image-analysis methods.[41][42][75][76]
First test and score details
First test
Proposed first test: Perturb a public or simulated optical perfusion dataset with motion and photon-count variation; compare error flagging at a fixed false-alarm rate.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 5/15; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). Preserved original co-membership proposal in o05. An added candidate, Shahar Alon (10/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: A physiologically meaningful reference is needed before interpreting calibrated signal estimates as clinical performance. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Original initiative conditions remain: o05 Nisan Ozana: Conditional role: assess readout feasibility only if a defined optical/acoustic observable maps to the target functional signal in the chosen culture model. o05 Jacob Goldberger: Conditional later role: evaluate uncertainty after a reference dataset and transparent signal-analysis baseline exist.
Connection 6
Ori Ernst
Original proposal
Proposal hypothesis: Ernst supplies proposition clusters and Goldberger supplies statistical language modelling and uncertainty estimation to decide which summary statements need review.[75][76][93][94][95]
First test and score details
First test
Proposed first test: On an adjudicated open corpus, compare uncertainty-based review with random review at the same budget; measure unsupported statements missed.
Score components
- complementarity
- 2
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 6/15; fit 9/10 (4 topic overlap + 2 complementarity + 3 feasible first test). Preserved original co-membership proposal in o01. An added candidate, Shahar Alon (10/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: Strong NLP overlap, with partially overlapping modelling roles rather than a new measurement modality. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 7
Orr Levy
Proposal hypothesis: Levy can define single-cell and microbiome prediction targets while Goldberger evaluates calibration under dataset shift.[5][6][75][76]
First test and score details
First test
Proposed first test: Hold out one open biological cohort and compare calibrated predictions with a simple feature baseline on error and abstention.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 7/15; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Cohort associations and predictive uncertainty are not evidence of ageing interventions. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 8
Gonen Singer
Proposal hypothesis: Goldberger's uncertainty estimates can inform Singer's cost-sensitive allocation of scarce review or service capacity.[75][76][79][80]
First test and score details
First test
Proposed first test: On a public classification benchmark with synthetic service costs, compare uncertainty-based and cost-based triage on missed errors and total cost.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 8/15; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Synthetic cost weights are not clinical or institutional policy values. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 9
Izack Cohen
Proposal hypothesis: Goldberger's statistical uncertainty can propagate uncertain demand predictions into Izack Cohen's queueing decisions.[75][76][97][98]
First test and score details
First test
Proposed first test: Compare plug-in demand forecasts with interval-aware queue policies on simulated waiting-time tails and capacity violations.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 9/15; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. The forecasting model and queue assumptions must be calibrated separately. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 10
Ilan Reuven Cohen
Proposal hypothesis: Goldberger can calibrate prediction uncertainty while Ilan Cohen chooses which classifications receive scarce resources.[75][76][119][120]
First test and score details
First test
Proposed first test: Compare fixed-threshold and budget-aware review on a public classification set with synthetic review costs; report missed errors and resource use.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 10/15; fit 9/10 (3 topic overlap + 3 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. The cost model and fairness constraints are hypothetical and need stakeholder definition. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work.
Connection 11
Yoli Shavit
Proposal hypothesis: Goldberger's multimodal uncertainty methods and Shavit's representation/generalization work share a direct calibration problem across domains.[75][76][127][128]
First test and score details
First test
Proposed first test: Compare two uncertainty approaches on a public image-plus-metadata task under domain shift, keeping training data and model size fixed.
Score components
- complementarity
- 2
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 11/15; fit 9/10 (4 topic overlap + 2 complementarity + 3 feasible first test). New pairing outside the frozen portfolio co-member graph. Strong topical overlap comes with redundant methodological capacity, so complementarity is not maximal. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Yoli Shavit biological/multimodal programme evidence is a 2025 announcement; current laboratory operation and biological experimental capability are not established.
Connection 12
Orit Shefi
Original proposal
Proposal hypothesis: Shefi supplies interpretable culture structure and perturbation labels while Goldberger can test uncertainty in image-derived growth measurements.[8][75][76]
First test and score details
First test
Proposed first test: On an existing approved or openly licensed culture-image set, compare a transparent morphology baseline with a calibrated model under illumination shifts.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 12/15; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o05. An added candidate, Shahar Alon (10/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: A reference dataset and image-level ground truth are prerequisites; do not equate predicted morphology with neuronal function. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Original initiative conditions remain: o05 Jacob Goldberger: Conditional later role: evaluate uncertainty after a reference dataset and transparent signal-analysis baseline exist.
Connection 13
Hila Chalutz-Ben Gal
Original proposal
Proposal hypothesis: Goldberger's uncertainty estimates can provide controlled confidence displays for Chalutz-Ben Gal's research on how AI changes professional decisions.[22][23][75][76]
First test and score details
First test
Proposed first test: Design an approved-study protocol using synthetic advice with calibrated versus misleading confidence; pilot a rubric for error detection and inappropriate deference.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 13/15; fit 7/10 (2 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o01. An added candidate, Shahar Alon (10/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: The shared problem is decision use, not a demonstrated joint research programme; participant access is unconfirmed. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Original initiative conditions remain: o01 Hila Chalutz-Ben Gal: Conditional later role: design a separately approved researcher-use study of noticing and correcting unsupported recommendations.
Connection 14
Mor Weiss
Original proposal
Proposal hypothesis: Goldberger can define an uncertainty-based triage calculation and Mor Weiss can examine private verification of that calculation, leaving empirical calibration to a separate evaluation.[75][76][123]
First test and score details
First test
Proposed first test: Encode a small threshold-based triage rule over synthetic uncertainty values and measure proof/verifier cost relative to direct evaluation.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 1
Why this rank
Rank 14/15; fit 6/10 (1 topic overlap + 3 complementarity + 2 feasible first test). Preserved original co-membership proposal in o01. An added candidate, Shahar Alon (10/10), ranks above this original because its stated pair-specific roles and first test score higher; this original is limited as follows: Cryptographic correctness cannot certify calibration or diagnostic validity, so the original connection remains exploratory. No automatic score boost for original membership. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Original initiative conditions remain: o01 Mor Weiss: Conditional later role: only after a protected-input threat model exists, specify a computation or access-policy statement that can be proven without revealing inputs.
Connection 15
Alexandra Simonovsky
Identity check needed: Conditional proposal: confirm the researcher identity and research interests before assessing this match.
Proposal hypothesis: Conditional on confirmation, Simonovsky's event-coordination experience could support a prototype clinic with uncertainty-evaluation tasks designed by Goldberger.[16][17][75][76]
First test and score details
First test
Conditional proposed first test: After identity confirmation, prepare synthetic prediction examples and test whether a common submission form captures errors and confidence consistently.
Score components
- complementarity
- 2
- feasible first test
- 2
- topic overlap
- 1
Why this rank
Rank 15/15; fit 5/10 (1 topic overlap + 2 complementarity + 2 feasible first test). New pairing outside the frozen portfolio co-member graph. Goldberger owns the scientific evaluation; no machine-learning specialization is inferred for Simonovsky. Equal scores use existing-first, then stable researcher ID.
Conditions
Analyst proposal hypothesis, not an established collaboration, commitment, evidence-confidence rating or assessment of researcher quality; confirm participation and access before work. Conditional: confirm Alexandra Simonovsky identity and current role before any team assignment. Evidence supports AI-agent event coordination only; research appointment, teaching role and technical research expertise are not verified. Excluded from confirmed-team claims.
External connections
10 candidates
Connection 1
Yarin Gal
University of Oxford
Original proposal
His Bayesian deep-learning, uncertainty, and AI-safety work could complement calibrated decision rules; no willingness is asserted.[75][76][78][452]
First test and score details
First test
Replay a held-out domain-shift benchmark and test whether the triage rule captures failures at a fixed review budget.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 1/10 after semantic revision; analyst score 10 = max(1, 4+3+3): topic overlap 4/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked. Original remains first under these components; original status breaks equal-score ties only, without a prestige bonus.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified.
Connection 2
Gregory Wornell
Massachusetts Institute of Technology
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Gregory Wornell's documented signal processing, statistical inference for calibrated evidence triage for clinical and research ai. The specific contribution is joint statistical inference and information constraints; this transfer is an analyst hypothesis.[75][76][110][503]
First test and score details
First test
Compare full-data inference with task-specific compressed statistics at fixed communication or storage budget using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare estimation error, calibration and bits per valid decision with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 4
Why this rank
Rank 2/10 after semantic revision; analyst score 10 = max(1, 4+3+3): topic overlap 4/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified.
Connection 3
Michael Bronstein
University of Oxford
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Michael Bronstein's documented geometric deep learning, graph neural networks for calibrated evidence triage for clinical and research ai. The specific contribution is geometric structure and graph representations; this transfer is an analyst hypothesis.[74][75][76][442]
First test and score details
First test
Compare a geometry/graph-aware model with a parameter-matched unstructured baseline under structural perturbations using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare held-out error, symmetry consistency and robustness to altered graph topology with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 3/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The refreshed Oxford profile also lists an Aithyra scientific-director role; the original Oxford institution string is preserved.
Show 7 moreShow fewer external connections
Connection 4
Dan Jurafsky
Stanford University
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Dan Jurafsky's documented natural-language processing, language and society for calibrated evidence triage for clinical and research ai. The specific contribution is language behaviour and semantic alignment; this transfer is an analyst hypothesis.[75][76][464]
First test and score details
First test
Perturb wording, social framing and answer incentives while keeping underlying evidence fixed using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare semantic consistency, sycophancy and subgroup differences with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 4/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. The official page lists sabbatical during 2026-2027; availability is not inferred.
Connection 5
Mirella Lapata
University of Edinburgh
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Mirella Lapata's documented natural-language processing, long-context understanding for calibrated evidence triage for clinical and research ai. The specific contribution is long-context evidence synthesis; this transfer is an analyst hypothesis.[75][76][96][467]
First test and score details
First test
Compare extractive and generative synthesis on a fixed multi-document set with sentence-to-source labels using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare supported-claim recall, contradiction handling and unsupported statements with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 5/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified.
Connection 6
Christopher Manning
Stanford University
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Christopher Manning's documented natural-language inference, summarization for calibrated evidence triage for clinical and research ai. The specific contribution is natural-language entailment and retrieval; this transfer is an analyst hypothesis.[75][76][472]
First test and score details
First test
Compare retrieved evidence with generated assertions on an annotated set of paraphrases and contradictions using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare entailment precision, retrieval recall and unsupported-claim rate with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 6/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified.
Connection 7
Andrea Montanari
Stanford University
Proposed capability match: Jacob Goldberger's deep learning, medical image analysis can be paired with Andrea Montanari's documented high-dimensional statistics, posterior sampling for calibrated evidence triage for clinical and research ai. The specific contribution is high-dimensional statistical baselines; this transfer is an analyst hypothesis.[75][76][479]
First test and score details
First test
Compare a regularized low-complexity estimator with a flexible model while varying sample size and dimensionality using a de-identified public multimodal evidence-triage dataset split by institution or acquisition condition. Compare generalization error, calibration and the sample-size threshold with uncalibrated single-model triage with the same evidence inputs.
Score components
- complementarity
- 3
- feasible first test
- 3
- topic overlap
- 3
Why this rank
Rank 7/10 after semantic revision; analyst score 9 = max(1, 3+3+3): topic overlap 3/4, complementarity 3/3, feasible first test 3/3. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified.
Connection 8
Caroline Uhler
Massachusetts Institute of Technology
Proposed capability match for Jacob Goldberger with Caroline Uhler: Causal methods can examine acquisition-induced triage error under a declared intervention model; de-identified observational records alone do not establish intervention effects.[75][76][129][497]
First test and score details
First test
Build a synthetic triage causal model with true case state, acquisition site, noisy modalities and an explicit intervention on sensor quality. Fit associational and causal-model predictors on the same observational records; compare predicted changes in triage error with simulator truth under held-out quality interventions. Treat feature deletion only as a sensitivity analysis, not proof of causation.
Score components
- complementarity
- 3
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 8/10 after semantic revision; analyst score 7 = max(1, 2+3+2): topic overlap 2/4, complementarity 3/3, feasible first test 2/3. Causal methods can examine acquisition-induced triage error under a declared intervention model; de-identified observational records alone do not establish intervention effects. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. Post-review scope: Causal methods can examine acquisition-induced triage error under a declared intervention model; de-identified observational records alone do not establish intervention effects. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Connection 9
Susan Athey
Stanford University
Proposed capability match for Jacob Goldberger with Susan Athey: Causal workflow evaluation is possible once an interface treatment and common decision outcome are specified; synthetic reviewer behavior cannot establish real clinical or institutional benefit.[75][76][433]
First test and score details
First test
Define treatment as showing a calibrated uncertainty flag and outcome as correct routing minus a fixed review-cost penalty. Generate synthetic cases and reviewer responses with a stipulated effect, randomize flag visibility, and compare pooled and stratified effect estimates against planted effects. Real human response and deployment effects require a later approved study.
Score components
- complementarity
- 2
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 9/10 after semantic revision; analyst score 6 = max(1, 2+2+2): topic overlap 2/4, complementarity 2/3, feasible first test 2/3. Causal workflow evaluation is possible once an interface treatment and common decision outcome are specified; synthetic reviewer behavior cannot establish real clinical or institutional benefit. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. Post-review scope: Causal workflow evaluation is possible once an interface treatment and common decision outcome are specified; synthetic reviewer behavior cannot establish real clinical or institutional benefit. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Connection 10
Michael Mitzenmacher
Harvard University
Proposed capability match for Jacob Goldberger with Michael Mitzenmacher: Scheduling contributes to triage service performance through a defined queue model; latency gains are separate from improving the evidence classifier.[75][76][101][478]
First test and score details
First test
Convert synthetic triage cases into arrivals with explicit urgency, service time and a fixed reviewer capacity. Compare randomized load balancing with FIFO routing under identical case labels and review accuracy; report queue delay, deadline misses and cost-weighted incorrect or late decisions. Keep the triage predictor unchanged.
Score components
- complementarity
- 2
- feasible first test
- 2
- topic overlap
- 2
Why this rank
Rank 10/10 after semantic revision; analyst score 6 = max(1, 2+2+2): topic overlap 2/4, complementarity 2/3, feasible first test 2/3. Scheduling contributes to triage service performance through a defined queue model; latency gains are separate from improving the evidence classifier. A bounded offline comparison is specified; required datasets and domain assumptions must still be checked.
Conditions
Proposed fit, not an assertion of a new or active relationship. Independent review pending; forecast confidence low. Partner interest, capacity, data access and any required experimental approvals/resources are unverified. Post-review scope: Scheduling contributes to triage service performance through a defined queue model; latency gains are separate from improving the evidence classifier. This revised proposal awaits independent targeted re-review; simulated outcomes would establish model behavior only, not biological, clinical or deployed benefit.
Evidence & open questions
- Clinical benefit and prospective workflow performance are unverified.
Inspect claim ratings and independent review
Jacob Goldberger has source-grounded capabilities in statistical machine learning and multimodal inference, represented here by deep learning, medical image analysis, speech and audio modeling. [75][76]
Moderate confidenceReview: reviewedThe sources establish public professional activity, not comparative quality, future performance, or willingness to participate.
Review record
- profiles_a: supports. Goldberger's official page and 2025 CV directly document statistical and deep learning across vision, medical imaging, speech, audio and language, including current calibration-related work.
Hypothesis for 2027–2031: Jacob Goldberger could explore calibrated evidence triage for clinical and research ai through the bounded first test described in this profile. [75][78]
Low confidenceReview: reviewedRetrospective calibration may not transfer to prospective clinical use.
Review record
- profiles_a: supports. The proposal links Goldberger's medical-imaging and calibration work with Gal's documented Bayesian uncertainty and robustness research. It is properly framed as a retrospective feasibility hypothesis rather than clinical efficacy.