FlowingFlowing
← Back to all development cases
This case was judged by an anonymized AI reviewer and has not been independently reviewed by people; inspect the draft, cards, and both outputs to assess it.
AutocompleteWith cards better14 / 280 · beu-706d1ae2e7ad7742

Predictive Coding: a Theoretical and Experimental Review

Quantitative biology · 2107.12979v4

FLOWING EVIDENCE BENCHMARK

How do we compare writing with and without cards?

At the same writing position, the same model produces one first result with cards and one without. All 280 pairs enter the overall anonymized review, which assesses whether a decision not to continue fits the draft.

Overall: 138 with-card wins, 26 ties, 107 no-card wins, 5 pairs where neither was usable, and 4 inconclusive judgments. Both sides wrote text in 218 pairs; 48 had a deliberate stop and 14 included an invalid output.

SAME DRAFT · WITH OR WITHOUT CARDS

01 · FIXED WRITING POSITION

MANUSCRIPT

Same writing position ▌

Matched autocomplete at the same draft position

02 · TWO MATCHED VERSIONS

SHARED BY BOTH

Draft, writing task, and generation model

A · With cards

Relevant literature cards supplied

B · Without cards

No literature cards supplied

Generation model

Same model and version

A → first output

B → first output

Anonymized AI judge

First outputs are anonymized as X and Y

Continuation review: accuracy, fit to the writing task and usability
Returns: X preferred / tie / Y preferred / both unusable

Review again with the answer order swapped

HOW DOES BLIND JUDGING WORK?

① Anonymize both outputs
The AI judge sees the same draft and both first outputs without knowing which received cards.

② Compare and swap order
The judge applies task-specific criteria in X/Y and then Y/X order.

③ Review disagreements
A third pass resolves disagreements. Incomplete reviews remain in the denominator.

How are positions and cards selected?

We selected 40 writing positions in each of seven fields, for 280 comparisons. The position and relevant literature cards were set before each comparison, so both versions began from the same draft.

The main result retains every frozen position, including places where cards might not add a useful next move. This avoids selecting cases based on their outputs, but 49% describes the whole cohort, not a win rate where evidence clearly had an opportunity to help. This cohort did not record that opportunity label for each position before generation, so those two groups cannot yet be separated reliably.

CARD USE · 140/280

Reported card use in continuations with text

The model reports whether a card affected its continuation. In 140 of 280 tests it produced continuation text and reported card use. The previous count of 150 included 10 cases without continuation text and has been corrected.

COMPARISON · 70 / 7 / 30

Which written version was better?

Of the 107 reported-use cases where both versions produced text, the blinded AI judge preferred the with-card version 70 times, found 7 about the same, and preferred the without-card version 30 times.

How are the two table percentages calculated?

ADOPTION COVERAGE

140reported use with text
÷
280all positions
=
50%

WITH-CARD VERSION PREFERRED

70with cards better
÷
107comparable tests
=
65%

The percentages use different groups: 50% counts continuations with text reporting card use among all 280 tests; 65% covers the 107 such cases with text on both sides. The generating model reports card use, while an anonymized AI judge compares writing quality.

Card use is self-reported by the generator. The anonymized AI judge sees the draft and source snippets, so source-aware preference remains possible; there was no independent human review. The overall result describes these 280 frozen positions, not a win probability for any writing task.

Manuscript writing position

Text excerpt · not a PDF page

Research manuscript · excerpt

Predictive Coding: a Theoretical and Experimental Review

1 Introduction

…A.K. Seth (2014) ) . Cybernetics as a science is focused on understanding the dynamics of interacting feedback loops for perception and control, based especially around the concept of error minimization. Control and filtering theory have, in a related but distinct way, been based around methods to minimize residual errors in both perception or action according to some objective for decades. As we shall see, standard methods such as Kalman Filtering ( Kalman (1960) ) or PID control Johnson & Moradi (2005) can be shown as special cases of predictive coding under certain restrictive assumptions.

…imizes the redundancy in the representation of the visual information closely resembles the centre-surround receptive fields which are well established in retinal ganglion cells ( Huang & Rao (2011) ) . This predictive coding approach was also applied to coding in the lateral geniculate nucleus (LGN), the thalamic structure that retinal signals pass through en-route to cortex, which was hypothesised to help remove temporal correlations in the input by subtracting out the retinal signal at previous timesteps using recurrent lateral inhibitory connectivity ( Huang & Rao (2011) ; Marino (2020) )

Mumford (1992) was perhaps the first to extend this theory of the retina and the LGN to a fully-fledged general theory of cortical function. His theory was motivated by simple observations about the

↓ Next paragraph was generated here

The highlighted text is near the writing position. This is a layout of the evaluation input, not a PDF screenshot.

A matched comparison at the same writing positionOverall AI review: With cards better

Both arms use the same model, task and manuscript context. Only the retrieved passages differ. The first recorded output from each arm is shown below.

Read overall AI review rationale

A = without cards · B = with cards

Both outputs complete the sentence about Mumford's motivating observations, but B is more precise and historically accurate. The draft specifically mentions Mumford (1992) extending a theory of 'the retina and the LGN' to cortical function. Output B correctly identifies the feedback connections as being 'from the cortex to the thalamus' (which includes the LGN), directly continuing the retinogeniculate thread. Output A's 'higher to lower cortical areas' is more generic and misses the specific thalamocortical/corticothalamic loop that Mumford's theory emphasized. B also adds the explanatory clause about feedback being 'too numerous to be explained by a purely feedforward model,' which captures the core theoretical motivation—this is the actual substance of Mumford's argument about why feedback must be doing predictive coding. A's 'feedforward connections' contrast is less precise than B's 'purely feedforward model of cortical processing.' B fits the specific argument trajectory better: retina→LGN→cortex with massive feedback to thalamus, not just generic cortico-cortical feedback.

Evidence

Same model and position · retrieved passages supplied · first output

Preferred by blind review

anatomy of the visual system, in particular the massive number of feedback connections from the cortex to the thalamus, which he argued were too numerous to be explained by a purely feedforward model of cortical processing.

Passages supplied to the Evidence version

Card 1 · Efficient Visual Coding: From Retina To V2

The human visual system has a hierarchical structure consisting of layers of processing, such as the retina, V1, V2, etc. Understanding the functional roles of these visual processing layers would help to integrate the psychophysiological and neurophysiological models into a consistent theory of hum…

Read full passage excerpt

The human visual system has a hierarchical structure consisting of layers of processing, such as the retina, V1, V2, etc. Understanding the functional roles of these visual processing layers would help to integrate the psychophysiological and neurophysiological models into a consistent theory of human vision, and would also provide insights to computer vision research. One classical theory of the early visual pathway hypothesizes that it serves to capture the statistical structure of the visual inputs by efficiently coding the visual information in its outputs. Until recently, most computational models following this theory have focused upon explaining the receptive field properties of one or two visual layers. Recent work in deep networks has eliminated this concern, however, there is till the retinal layer to consider. Here we improve on a previously-described hierarchical model Recursive ICA (RICA) [1] which starts with PCA, followed by a layer of sparse coding or ICA, followed by a …

Card 2 · Time-Causal and Time-Recursive Spatio-Temporal Receptive Fields

We present an improved model and theory for time-causal and time-recursive spatio-temporal receptive fields, obtained by a combination of Gaussian receptive fields over the spatial domain and first-order integrators or equivalently truncated exponential filters coupled in cascade over the temporal d…

Read full passage excerpt

We present an improved model and theory for time-causal and time-recursive spatio-temporal receptive fields, obtained by a combination of Gaussian receptive fields over the spatial domain and first-order integrators or equivalently truncated exponential filters coupled in cascade over the temporal domain. Compared to previous spatio-temporal scale-space formulations in terms of non-enhancement of local extrema or scale invariance, these receptive fields are based on different scale-space axiomatics over time by ensuring non-creation of new local extrema or zero-crossings with increasing temporal scale. Specifically, extensions are presented about (i) parameterizing the intermediate temporal scale levels, (ii) analysing the resulting temporal dynamics, (iii) transferring the theory to a discrete implementation in terms of recursive filters over time, (iv) computing scale-normalized spatio-temporal derivative expressions for spatio-temporal feature detection and (v) computational modelli …

Card 3 · Modeling the development of coarse-to-fine processing in the central visual pathway

The sequential analysis of information in a coarse-to-fine manner is a fundamental mode of processing in the visual pathway. Spatial frequency (SF) tuning, arguably the most fundamental feature of spatial vision, provides particular intuition within the coarse-to-fine framework: low spatial frequenc…

Read full passage excerpt

The sequential analysis of information in a coarse-to-fine manner is a fundamental mode of processing in the visual pathway. Spatial frequency (SF) tuning, arguably the most fundamental feature of spatial vision, provides particular intuition within the coarse-to-fine framework: low spatial frequencies convey global information about an image (e.g., general orientation), while high spatial frequencies carry more detailed information (e.g., edges). In this paper, we study the development of cortical spatial frequency tuning. As feedforward input from the lateral geniculate nucleus (LGN) has been shown to have significant influence on cortical coarse-to-fine processing, we present a firing-rate based thalamocortical model which includes both feedforward and feedback components. We analyze the relationship between various model parameters (including cortical feedback strength) and responses. We confirm the importance of the antagonistic relationship between the center and surround respons …

Card 4 · The Dynamics of Image Processing by Feature Maps in the Primary Visual Cortex

The operational characteristics of a linear neural network image processing system based on the brain's vision system are investigated. The final stage of the network consists of edge detectors of various orienations arranged in a feature map, corresponding to the primary visual cortex, or V1. The l…

Read full passage excerpt

The operational characteristics of a linear neural network image processing system based on the brain's vision system are investigated. The final stage of the network consists of edge detectors of various orienations arranged in a feature map, corresponding to the primary visual cortex, or V1. The lateral geniculate nucleus is modeled as a preprocessing stage. Excitatory forward and inhibitory backward connections exist between the LGN and V1. By a method of reconstructing the input images in terms of V1 activities, the simulations show that images can be faithfully represented in V1 by the proposed network. The signal-to-noise ratio of the image is improved by the representation, and compression ratios of well over two-hundred are possible. Lateral interacations between V1 neurons sharpen their orientational tuning. We further study the dynamics of the processing, showing that the rate of decrease of the error of the representation is maximized for the receptive fields used, and we de …

Card 5 · Receptive Field and Feature Map Formation in the Primary Visual Cortex via Hebbian Learning with Inhibitory Feedback

A linear neural network is proposed for mamalian vision system in which backward connections from the primary visual cortex (V1) to the lateral geniculate nucleus play a key role. The backward connections control the flow of information from the LGN to V1 in such a way as to maximize the rate of tra…

Read full passage excerpt

A linear neural network is proposed for mamalian vision system in which backward connections from the primary visual cortex (V1) to the lateral geniculate nucleus play a key role. The backward connections control the flow of information from the LGN to V1 in such a way as to maximize the rate of transfer of information from the LGN to V1. The application of hebbian learning to the forward and backward connections causes the formation of receptive fields which are sensitive to edges, bars, and spatial frequencies of preferred orientations. Receptive field types in V1 are shown to depend on the density of the afferent connections in the LGN. Orientational preferences are organized in the primary visual cortex by the application of lateral interactions during the learning phase. Change in the size of the eye between the immature and mature animal is shown be an important factor in the development of V1 organization. The orgainization of the mature network is compared to that found in the …

Model-reported card use

The model did not report that any card materially affected this completion.

No Evidence

Same model and position · no retrieved passages · first output

Not preferred in this comparison

anatomy of the cortex, in particular the massive number of feedback connections from higher to lower cortical areas relative to the feedforward connections.

Earlier card-use comparison

The earlier diagnostic compared only cases with reported card use and text on both sides; its scope differs from the overall AI review above.

This case does not have both with-card continuation text and a model-reported use record. It remains among all 280 tests but does not enter the reported-use comparison.

The passages are input material; their presence does not mean an output used them or that they support every claim. Draft and source passages are excerpted; no simulated scores or PDF appear here.