Triple

T10520598
Position Surface form Disambiguated ID Type / Status
Subject Ronald J. Williams E248157 entity
Predicate coAuthorOf P2389 FINISHED
Object “Simple statistical gradient-following algorithms for connectionist reinforcement learning” E426681 NE FINISHED

How this triple was built (2 steps)

Every LLM step that produced this triple, in pipeline order — named-entity classification, the disambiguation choices (the exact options shown, with the pick highlighted), and the generated description. The batch + timestamp of each is in the Provenance table below.

NER Named-entity recognition gpt-5-mini
Instruction
Given a phrase, classify it is english named entity (e.g., persons, organizations, works of art) in Latin script, or not (e.g., literals, dates, URLs, verbose phrases). For disambiguation, the statement where the phrase occurs as object is also given. Please return a JSON object with `phrase` (string, the phrase being analyzed) and `is_ne` (boolean, indicating whether the phrase is a Named Entity).
Input
Phrase: “Simple statistical gradient-following algorithms for connectionist reinforcement learning” | Statement: [Ronald J. Williams, coAuthorOf, “Simple statistical gradient-following algorithms for connectionist reinforcement learning”]
NED1 Entity disambiguation (via context triple) gpt-5-mini-2025-08-07 
Target entity: “Simple statistical gradient-following algorithms for connectionist reinforcement learning”
Context triple: [Ronald J. Williams, coAuthorOf, “Simple statistical gradient-following algorithms for connectionist reinforcement learning”]
  • A. “Learning representations by back-propagating errors”
    “Learning representations by back-propagating errors” is a landmark 1986 research paper that popularized the backpropagation algorithm for training multi-layer neural networks, helping to launch the modern field of deep learning.
  • B. Q-learning
    Q-learning is a model-free reinforcement learning algorithm that learns an action-value function to optimize decision-making by estimating the expected cumulative reward for each state-action pair.
  • C. Deep Q-Learning
    Deep Q-Learning is a reinforcement learning algorithm that uses deep neural networks to approximate Q-values, enabling agents to learn effective policies directly from high-dimensional inputs like raw images.
  • D. REINFORCE chosen
    REINFORCE is a classic Monte Carlo policy gradient algorithm in reinforcement learning that optimizes stochastic policies by estimating gradients from sampled returns.
  • E. Hebbian learning
    Hebbian learning is a neurobiological and computational learning principle often summarized as "cells that fire together wire together," where the connection between neurons is strengthened when they are activated simultaneously.
  • F. None of above.
  • G. Unsure - the case is ambiguous/there is not enough information to decide.

Provenance (3 batches)

The batch behind each pipeline step, in order, with when it ran. Timestamps are batch-level — stages were processed in waves, so the object chain (NER → NED1 → NEDg → NED2) reads in order, but predicate / elicitation batches can sit in a different wave.

Step Stage Batch ID Status When
creating Elicitation batch_69d381c4aa948190942e1d803143fb0e completed April 6, 2026, 9:49 a.m.
NER Named-entity recognition batch_69d509de0b3081909bec337aa8ff193e completed April 7, 2026, 1:42 p.m.
NED1 Entity disambiguation (via context triple) batch_69d90e119fe4819085e5c1c6e71e6260 completed April 10, 2026, 2:49 p.m.
Created at: April 6, 2026, 12:28 p.m.