The DOI matches its Crossref record; 83 of the 95 references it cites were found in a registry or an index; 11 of the 31 sources its claims cite were checked for support.
Design and evaluation of a global workspace agent embodied in a realistic multimodal environment
Open accessAbstract
As the apparent intelligence of artificial neural networks (ANNs) advances, they are increasingly likened to the functional networks and information processing capabilities of the human brain. Such comparisons have typically focused on particular modalities, such as vision or language. The next frontier is to use the latest advances in ANNs to design and investigate scalable models of higher-level cognitive processes, such as conscious information access, which have historically lacked concrete and specific hypotheses for scientific evaluation. In this work, we propose and then empirically assess an embodied agent with a structure based on global workspace theory (GWT) as specified in the recently proposed “indicator properties” of consciousness. In contrast to prior works on GWT which utilized single modalities, our agent is trained to navigate 3D environments based on realistic audiovisual inputs. We find that the global workspace architecture performs better and more robustly at smaller working memory sizes, as compared to a standard recurrent architecture. Beyond performance, we perform a series of analyses on the learned representations of our architecture and share findings that point to task complexity and regularization being essential for feature learning and the development of meaningful attentional patterns within the workspace.
Overview
CoChat Summary
TL;DR
The authors design an embodied global-workspace agent satisfying GWT indicator properties and train it on realistic audiovisual 3D navigation. It outperforms a GRU baseline mainly at smaller working-memory sizes, while larger workspaces yield little benefit and attention patterns become less dynamic.
Key findings
- The global workspace agent achieved ~80% success across all working-memory sizes apart from size 256, where some runs performed poorly and were considered possible outliers; the baseline GRU achieved only over 60% across all memory sizes and became competitive only at higher sizes. For reference, a random policy achieved 0% and the original SAVi RL agent reached 25% on average.
- Ablating working-memory size from 32 to 512 showed negligible performance differences for the global workspace agent, suggesting the task’s short-term memory demand is small and that SAVi is not an adequate benchmark for working-memory capacity; larger agents did not perform significantly better.
- Attention analyses showed all agents primarily used cross-attention across input modalities; smaller, more bottlenecked agents developed more mixed attention patterns integrating modalities over time, whereas larger workspaces produced more binary attention that saturated early, with working memory and null keys largely unattended.
- Probing found acoustic features encoded target object category and visual features encoded scene; target category from visual features ranged 14–37% versus chance 4%, scene from audio ranged 6–14% versus chance 1.79%, and after acoustic cue cutoff target-category probing accuracy dropped abruptly for both variants, though for memory sizes ≤128 global workspace agents had higher average probing accuracy than GRU counterparts. Broadcast weight analysis indicated previous working memory was prioritized over current sensory inputs in both audio and vision encoders.
Methods
- Architecture: an embodied agent with independent visual and acoustic CNN+GRU encoders, a recurrent working memory/global workspace, cross-attention bottleneck with a null input, and global broadcast from previous working memory to encoders; the baseline GRU ablates global broadcast and cross-attention while retaining independent encoders and recurrence.
- Task and data: SoundSpaces/SAVi semantic audiovisual navigation in 3D scanned real environments, with 21 target object categories and 56 training scenes; observations were 128×128×3 RGB images and 65×25×2 spectrograms, with four discrete actions (forward, turn left, turn right, stop).
- Training and analysis: behavioral cloning on 500,000 oracle tuples, batch size 10 contiguous trajectories of length 150, entropy coefficient 0.2, Bayesian hyperparameter optimization, final five seeds for 50,000 iterations across memory sizes 32, 64, 128, 256, and 512; each run took ~5 days on an NVIDIA RTX 3090, requiring more than 8,616 GPU hours total. Evaluation used interquartile mean ±95% CI over the last 50 evaluations, probing MLPs, attention weights, and normalized input-weight magnitudes.
Limitations
- Stated: SAVi is not an adequate benchmark for testing working-memory capacity; a more sophisticated task or environment is needed; developing realistic simulator environments is labor-intensive and beyond the scope of this work. Performance sometimes declined after the halfway point, which may be an artifact of hyperparameter optimization runs stopping early to reduce computational requirements by ~4,000 GPU hours.
- Stated: neither the conventional GRU nor the proposed attention mechanism preserved high enough target-category probing accuracy after sound cutoff, contrasting with the hypothesized ability of the global workspace to sustain a representation over time; no consistent attention patterns appeared across runs, and all agents achieved similar task performance despite different attention.
- Evident: findings come from one audiovisual navigation task using imitation learning from an oracle rather than online RL; the work evaluates indicator properties, not consciousness itself; and the baseline retains some GWT/RPT properties (GWT-1, RPT-1/2), so the ablation is partial.
Model-written summary. Check it against the paper before relying on it.
Integrity signals
Registry
Cites
Statistics
Figures









Verification
Claims the paper makes
7 Valid · 4 Not valid · 12 Not checked · 63 Own results and interpretations
At least one cited source doesn’t support this claim or points to a different paper, and none confirmed it — see the verdicts below.
In the paper
“However, prior work has been restricted to unimodal input data, namely either visual or textual information (Goyal et al., 2021; Juliani et al., 2022b).”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.1 Functional theories of consciousness
What the sources say
- Disputedcites Goyal et al., 2021 sourceExperimental — accuracy not yet measured
The Experiments section (p. 8) describes a shared-workspace task whose input combines an image with a question. It says the model concatenates image patches with a representation of the question, contrary to the claim that this prior work was restricted to visual-only or textual-only input. (LLM, conf=0.90)
Checked by a reasoning model — may be wrong.
- Not matchedcites Juliani et al., 2022b
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Disputed
The cited sources that could be checked couldn’t confirm or dispute this claim.
In the paper
“While Transformer-based architectures have meteorically risen in popularity and become widely used across numerous academic and industrial applications, they lack the overall structure of a system with a global workspace.”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.2 Implementations of global workspace theory
What the sources say
- Could not verifycites Vaswani et al., 2017 sourceExperimental — accuracy not yet measured
The Model Architecture section (p. 2) describes the Transformer's encoder–decoder structure, and §3.2.3 (p. 5) describes how its attention layers operate. Neither assesses that structure against a global workspace; a search for “global workspace” found no such discussion in our copy. (LLM, conf=0.76)
- Not checked — no full text of the sourcecites Butlin et al., 2023 doi:10.48550/arxiv:2308.08708
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Could not verify
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“Despite these additions, the agent was only able to achieve an average success rate of 25%.”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.7 Agent training
What the sources say
- Partly supportedcites Chen et al., 2021 sourceExperimental — accuracy not yet measured
Our copy reports a success rate of ~25% for SAVi, but only on the unheard-sounds split: Section 7.2 (Implementation Details and Analysis of the Baselines) states 'our SAVi model 24.8% success rate and 17.2% SPL' on the unheard sound test split, and Table 1 (Sec. 4) lists 24.8 unheard / 33.9 heard. The claim's characterization as an 'average success rate of 25%' is not supported — the paper nowhere averages across settings, and averaging its two reported success rates would give ~29%, not 25%. So the magnitude traces to the source but the 'average' framing and the implied overall performance figure do not. (LLM, conf=0.72)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
The sources with a verdict support this claim; the rest weren’t checked yet, have no full text we could read, or their verdicts aren’t rated yet.
In the paper
“each of these previous studies have lacked evaluation of the model while embodied within a multimodal environment.”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 1 Introduction
What the sources say
- Supportedcites Goyal et al., 2021 source
The Experiments section (pp. 6, 8–9) describes evaluations on benchmark tasks, including image-and-question reasoning, video object tracking, physical prediction, and Atari games. Although some tasks involve more than one input type, the paper does not describe evaluating its model while embodied in a multimodal environment. (LLM, conf=0.84)
- Not matchedcites Juliani et al., 2022b
Couldn’t be matched to a paper in Crossref, so it can’t be checked.
- Not checked — no full text of the sourcecites Butlin et al., 2023 doi:10.48550/arxiv:2308.08708
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Supported
The cited sources support only part of this claim — read each source’s verdict below.
In the paper
“The SAVi task (Chen et al., 2021) features two semantic concepts that can be queried from the features learned by the agent.”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.8 Agent analysis
What the sources say
- Partly supportedcites Chen et al., 2021 sourceExperimental — accuracy not yet measured
Section 4.2 (page unavailable) says SAVi’s learned goal descriptor predicts two properties from acoustic observations: the target’s location and object category. It explicitly calls location spatial and object category semantic, so the paper does not support describing both as semantic concepts. (LLM, conf=0.86)
plain- Asserted: passes — The authors put it forward as true.
- Checkable: passes — It can be checked against evidence.
- Load-bearing: passes — The paper’s conclusions depend on it.
- Based: passes — The paper gives a reason why it holds.
- Partly supported
37 more claims · 1 with their cited sources checked
Create a free account to see all 86 claims and to verify the 3 references still open.
- It may state two separate claims as one. 26
- Probably not a substantial claim. 23
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
The cited source says what the claim attributes to it.
In the citing paper
“Researchers aiming to i mprove AI capabilities have proposed -and in some cases built -systems that intentionally reproduce c omputational features associated with human consciousness [ 4,5].”
From Identifying indicators of consciousness in AI systems (2026) · cites this paper as [5]
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
Each contradiction between two passages listed here was confirmed by an independent check.
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.2 Implementations of global workspace theory
“the Perceiver architecture notably lacks global broadcasting to the independent, pre-processing modules, thus falling short of satisfying GWT-2.”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.1 Functional theories of consciousness
“| GWT-3 | Global broadcast makes information in the workspace available to all modules |”
Rated minor: it would not touch a result — a value no result rests on, a p-value that stays on the same side of the significance level, or a conflict that is clerical, resolved elsewhere in the paper, or outside its own findings.
In the paper
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.7 Agent training
“We augment this with an entropy regularization term, scaled by a coefficient η = 0.2, to mitigate overfitting of the agent's policy”
§ Design and evaluation of a global workspace agent embodied in a realistic multimodal environment > 2 Materials and methods > 2.7 Agent training
“| Entropy coefficient | 0.497 | 0.066 | 0.221 | 0.055 | 0.363 | 0.420 | 0.223 | 0.396 | 0.200 | 0.331 |”
Citations
Cites
Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, L., & Polosukhin, I. (2017) · arXiv · doi:10.48550/arxiv.1706.03762 · Cited by 120,404
Very Deep Convolutional Networks for Large-Scale Image Recognition
Simonyan, K., & Zisserman, A. (2014) · arXiv · doi:10.48550/arxiv.1409.1556 · Cited by 114,148
ImageNet classification with deep convolutional neural networks
Krizhevsky, A., Sutskever, I., & Hinton, G. E. (2017) · Communications of the ACM · doi:10.1145/3065386 · Cited by 112,155
Hochreiter, S., & Schmidhuber, J. (1997) · Neural Computation · doi:10.1162/neco.1997.9.8.1735 · Cited by 102,318 · cited 3×
“Similarly, biological research on other aspects of cognition, such as memorization, has benefited from comparative studies between populations of neurons and their artificial analogs (Bedia et al., 2007; Li and Fan, 2019), such as recurrent neural networks (RNNs), which are a prominent class of architecture used to proc…”
BackgroundAdam: A Method for Stochastic Optimization
Kingma, D. P., & Ba, J. (2014) · arXiv · doi:10.48550/arxiv.1412.6980 · Cited by 98,946
Exploiting Generative AI to Scale up Intelligent Tutoring Systems
Jakubův, J., Chvalovský, K., Goertzel, Z., Kaliszyk, C., Olšák, M., Piotrowski, B., Schulz, S., Suda, M., & Urban, J. (2023) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/lipics.itp.2023.19 · Cited by 76,658
ImageNet Large Scale Visual Recognition Challenge
Russakovsky, O., Deng, J., Su, H., Krause, J., Satheesh, S., Ma, S., Huang, Z., Karpathy, A., Khosla, A., Bernstein, M., et al. (2015) · International Journal of Computer Vision · doi:10.1007/s11263-015-0816-y · Cited by 39,380 · cited 1×
“…(CNNs; LeCun and Bengio, 1995; Krizhevsky et al., 2012; Simonyan and Zisserman, 2014) in the ImageNet visual object recognition competition (Russakovsky et al., 2015) spurred comparisons between trained CNNs and areas of the brain related to visual processing (Afraz et al., 2014; Seijdel…”
MethodNeural Machine Translation by Jointly Learning to Align and Translate
Bahdanau, D., Cho, K., & Bengio, Y. (2014) · arXiv · doi:10.48550/arxiv.1409.0473 · Cited by 16,213
Parallel Distributed Processing: Explorations in the Microstructure of Cognition: Foundations
Rumelhart, D. E., McClelland, J. L., & AU (1986) · doi:10.7551/mitpress/5236.001.0001 · Cited by 15,686
Proceedings of the 24th international conference on Machine learning
(2007) · doi:10.1145/1273496 · Cited by 11,630
Proximal Policy Optimization Algorithms
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017) · arXiv · doi:10.48550/arxiv.1707.06347 · Cited by 10,625 · cited 1×
“…development of environments (Baker et al., 2019; Chen et al., 2020, 2021, 2022; Cobbe et al., 2020; Suarez et al., 2021) and agent architectures (Mnih et al., 2013; Schulman et al., 2017; Espeholt et al., 2018; Hafner et al., 2020) that learn to achieve a given goal while navigating through them.”
BackgroundPlaying Atari with Deep Reinforcement Learning
Mnih, V., Kavukcuoglu, K., Silver, D., Graves, A., Antonoglou, I., Wierstra, D., & Riedmiller, M. (2013) · arXiv · doi:10.48550/arxiv.1312.5602 · Cited by 9,471
Reinforcement Learning: An Introduction
Johnson, J. D., Li, J., & Chen, Z. (2000) · Neurocomputing · doi:10.1016/s0925-2312(00)00324-6 · Cited by 8,414
On the Properties of Neural Machine Translation: Encoder–Decoder Approaches
Cho, K., van Merrienboer, B., Bahdanau, D., & Bengio, Y. (2014) · Proceedings of SSST-8, Eighth Workshop on Syntax, Semantics and Structure in Statistical Translation · doi:10.3115/v1/w14-4012 · Cited by 7,099 · cited 4×
“Solving some complex tasks also requires the ability to store and process information in a working memory—a capacity which can be made possible through the use of RNNs (Hochreiter and Schmidhuber, 1997; Cho et al., 2014; Hafner et al., 2020).”
BackgroundMethodInfluentialFukushima, K. (1980) · Biological Cybernetics · doi:10.1007/bf00344251 · Cited by 5,759 · cited 1×
“While neuroscience had a profound influence on the fields of artificial neural networks (ANNs) and deep learning (DL) in the past (Rosenblatt, 1962; Fukushima, 1980; Rumelhart et al., 1986; LeCun and Bengio, 1995), in recent years the direction of influence has largely changed, and deep neural networks…”
BackgroundConvolutional networks for images, speech, and time series
LeCun, Y., & Bengio, Y. (1998) · HAL (Le Centre pour la Communication Scientifique Directe) · doi:10.5555/303568.303704 · Cited by 4,231
(1988) · Cited by 3,656
On a confusion about a function of consciousness
Block, N. (1995) · Behavioral and Brain Sciences · doi:10.1017/s0140525x00038188 · Cited by 3,364 · cited 1×
“GWT (Baars, 1993, 2005) is a framework proposed to formalize access consciousness—the idea that what is conscious is information that is accessible across various mental processes (Block, 1995).”
BackgroundLong short-term memory recurrent neural network architectures for large scale acoustic modeling
Sak, H., Senior, A., & Beaufays, F. (2014) · Interspeech 2014 · doi:10.21437/interspeech.2014-80 · Cited by 2,973 · cited 1×
“…studies between populations of neurons and their artificial analogs (Bedia et al., 2007; Li and Fan, 2019), such as recurrent neural networks (RNNs), which are a prominent class of architecture used to process temporal data (Hochreiter and Schmidhuber, 1997; Cho et al., 2014; Sak et al., 2014).”
BackgroundFacing up to the problem of consciousness
Chalmers, D. J. (1996) · Toward a Science of Consciousness · doi:10.7551/mitpress/6860.003.0003 · Cited by 1,966
Principles of Neurodynamics: Perceptrons and the Theory of Brain Mechanisms
White, B. W., & Rosenblatt, F. (1963) · The American Journal of Psychology · doi:10.2307/1419730 · Cited by 1,792
Proceedings for the Annual Meeting of the Cognitive Science Society
Dziemianko, M., Clarke, A. D. F., & Keller, F. (2013) · Cited by 1,726
PRINCIPLES OF NEURODYNAMICS. PERCEPTRONS AND THE THEORY OF BRAIN MECHANISMS
ROSENBLATT, F. (1961) · doi:10.21236/ad0256582 · Cited by 1,691
Embodiment and Cognitive Science
Gibbs, Jr. (2005) · Cambridge University Press eBooks · doi:10.1017/cbo9780511805844 · Cited by 1,589
ALVINN: An autonomous land vehicle in a neural network
Pomerleau, D. A. (2018) · Carnegie Mellon University · doi:10.1184/r1/6603146.v1 · Cited by 1,340
ALVINN, an autonomous land vehicle in a neural network
Pomerleau, D. (2018) · doi:10.1184/r1/6603146 · Cited by 1,338
Towards a true neural stance on consciousness
Lamme, V. A. F. (2006) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2006.09.001 · Cited by 1,149 · cited 1×
“…a comprehensive overview of various theories of consciousness from a functionalist perspective, including recurrent processing theory (RPT; Lamme, 2006, 2010), GWT (Baars, 1993, 2005), and AST (Graziano, 2017; Wilterson et al., 2020), while also touching upon the aspects of agency and…”
BackgroundOn the Properties of Neural Machine Translation: Encoder-Decoder ApproachesPreprint
Cho, K., van Merriënboer, B., Bahdanau, D., et al. (2014) · arXiv (Cornell University) · doi:10.48550/arxiv.1409.1259 · Cited by 1,112 · cited 4×
“Solving some complex tasks also requires the ability to store and process information in a working memory—a capacity which can be made possible through the use of RNNs (Hochreiter and Schmidhuber, 1997; Cho et al., 2014; Hafner et al., 2020).”
BackgroundMethodInfluentialStable-Baselines3: Reliable Reinforcement Learning Implementations
Raffin, A., Hill, A., Gleave, A., Kanervisto, A., Ernestus, M., & Dormann, N. (2021) · elib (German Aerospace Center) · Cited by 1,000
Global workspace theory of consciousness: toward a cognitive neuroscience of human experience
Baars, B. J. (2005) · Progress in Brain Research · doi:10.1016/s0079-6123(05)50004-9 · Cited by 888 · cited 8×
“This theory was heavily inspired by the structure of the biological brain: the existence of multiple specialized functional networks, and the fact that they process the information flow from the environment in parallel (Baars, 1993, 2005).”
BackgroundInfluentialEmpirical support for higher-order theories of conscious awareness
Lau, H., & Rosenthal, D. (2011) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2011.05.009 · Cited by 869 · cited 1×
“In addition, the global workspace can be interpreted as working memory within the brain, as it is also expected to maintain the information required to sustain a state of consciousness across variable lengths of time (Lau and Rosenthal, 2011; Park and Tallon-Baudry, 2014).”
BackgroundCognitron: A self-organizing multilayered neural network
Fukushima, K. (1975) · Biological Cybernetics · doi:10.1007/bf00342633 · Cited by 757 · cited 1×
“In practice, each probe is defined as a 2-layer MLP with ReLU activations (Fukushima, 1975).”
MethodSolving Rubik's Cube with a Robot Hand
OpenAI, Akkaya, I., Andrychowicz, M., Chociej, M., Litwin, M., McGrew, B., Petron, A., Paino, A., Plappert, M., Powell, G., et al. (2019) · arXiv · doi:10.48550/arxiv.1910.07113 · Cited by 602
Understanding intermediate layers using linear classifier probes
Alain, G., & Bengio, Y. (2016) · arXiv · doi:10.48550/arxiv.1610.01644 · Cited by 581
IMPALA: Scalable Distributed Deep-RL with Importance Weighted Actor-Learner Architectures
Espeholt, L., Soyer, H., Munos, R., Simonyan, K., Mnih, V., Ward, T., Doron, Y., Firoiu, V., Harley, T., Dunning, I., et al. (2018) · arXiv (Cornell University) · doi:10.48550/arxiv.1802.01561 · Cited by 553
How neuroscience will change our view on consciousness
Lamme, V. A. F. (2010) · Cognitive Neuroscience · doi:10.1080/17588921003731586 · Cited by 487 · cited 1×
“…a comprehensive overview of various theories of consciousness from a functionalist perspective, including recurrent processing theory (RPT; Lamme, 2006, 2010), GWT (Baars, 1993, 2005), and AST (Graziano, 2017; Wilterson et al., 2020), while also touching upon the aspects of agency and…”
BackgroundAnalysis Methods in Neural Language Processing: A Survey
Belinkov, Y., & Glass, J. (2019) · Transactions of the Association for Computational Linguistics · doi:10.1162/tacl_a_00254 · Cited by 477
Ba, J. L., Kiros, J., & Hinton, G. E. (2016) · arXiv (Cornell University) · doi:10.48550/arxiv.1607.06450 · Cited by 440
Perceiver: General Perception with Iterative Attention
Jaegle, A., Gimeno, F., Brock, A., Zisserman, A., Vinyals, O., & Carreira, J. (2021) · arXiv · doi:10.48550/arxiv.2103.03206 · Cited by 414
Brains and algorithms partially converge in natural language processing
Caucheteux, C., & King, J.-R. (2022) · Communications Biology · doi:10.1038/s42003-022-03036-1 · Cited by 318 · cited 1×
“For example, these models have been used as tools to explore and generate hypotheses on the neural mechanisms involved in language comprehension and production (Caucheteux and King, 2022).”
MethodInductive biases for deep learning of higher-level cognition
Goyal, A., & Bengio, Y. (2022) · Proceedings of the Royal Society A: Mathematical, Physical and Engineering Sciences · doi:10.1098/rspa.2021.0068 · Cited by 300 · cited 1×
“…for studying natural intelligence, we believe that now is a prime time to use the latest advances in DL to investigate higher-order thought processes and functions in the brain, such as the access and processing of conscious information (Bengio, 2017; Goyal and Bengio, 2022; Juliani et al., 2022a).”
MethodRecurrent Independent MechanismsPreprint
Goyal, A., Lamb, A., Hoffmann, J., Sodhani, S., Levine, S., Bengio, Y., & Schölkopf, B. (2019) · arXiv (Cornell University) · doi:10.48550/arxiv.1909.10893 · Cited by 262 · cited 1×
“To allow the working memory to be “inattentive”, a null input component was added to the key and value components of the attention mechanism, as proposed in RIMs (Goyal et al., 2019).”
MethodSingh, S. (2026) · SSRN Electronic Journal · doi:10.2139/ssrn.7475259 · Cited by 252
Mastering Atari with Discrete World Models
Hafner, D., Lillicrap, T., Norouzi, M., & Ba, J. (2020) · arXiv · doi:10.48550/arxiv.2010.02193 · Cited by 247
The neural subjective frame: from bodily signals to perceptual consciousness
Park, H.-D., & Tallon-Baudry, C. (2014) · Philosophical Transactions of the Royal Society B: Biological Sciences · doi:10.1098/rstb.2013.0208 · Cited by 246
SoundSpaces: Audio-Visual Navigation in 3D Environments
Chen, C., Jain, U., Schissler, C., Gari, S. V. A., Al-Halah, Z., Ithapu, V. K., Robinson, P., & Grauman, K. (2020) · Lecture Notes in Computer Science · doi:10.1007/978-3-030-58539-6_2 · Cited by 202 · cited 9×
“Altering the overall structure of the agent by having a joint input processing module over both the visual and acoustic modalities would introduce a significant gap when compared to the global workspace agent and the original reference implementation provided in SoundSpaces 1.0 (Chen et al., 2020).”
BackgroundMethodInfluentialPerceiver IO: A General Architecture for Structured Inputs & Outputs
Jaegle, A., Borgeaud, S., Alayrac, J.-B., Doersch, C., Ionescu, C., Ding, X. D., Koppula, S., Zoran, D., Brock, A., Shelhamer, E., et al. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2107.14795 · Cited by 198
A cognitive architecture that combines internal simulation with a global workspace
Shanahan, M. (2006) · Consciousness and Cognition · doi:10.1016/j.concog.2005.11.005 · Cited by 197 · cited 2×
“Early computational implementations of GWT were necessarily limited in their sophistication and what domains they could be applied to Baars (1993) and Shanahan (2006).”
BackgroundConsciousness in Artificial Intelligence: Insights from the Science of Consciousness
Butlin, P., Long, R., Elmoznino, E., Bengio, Y., Birch, J., Constant, A., Deane, G., Fleming, S. M., Frith, C., Ji, X., et al. (2023) · arXiv · doi:10.48550/arxiv.2308.08708 · Cited by 196
Self-citationLeveraging Procedural Generation to Benchmark Reinforcement Learning
Cobbe, K., Hesse, C., Hilton, J., & Schulman, J. (2019) · arXiv · doi:10.48550/arxiv.1912.01588 · Cited by 196
Policy Optimization with Demonstrations
Kang, B., Jie, Z., & Feng, J. (2018) · International Conference on Machine Learning · Cited by 195
SoundSpaces 2.0: A Simulation Platform for Visual-Acoustic Learning
Chen, C., Schissler, C., Garg, S., Kobernik, P., Clegg, A., Calamia, P. T., Batra, D., Robinson, P. W., & Grauman, K. L. (2022) · neural information processing systems · doi:10.52202/068431-0647 · Cited by 172
SoundSpaces 2.0: A Simulation Platform for Visual-Acoustic LearningPreprint
Chen, C., Schissler, C., Garg, S., et al. (2022) · Neural Information Processing Systems · doi:10.48550/arxiv.2206.08312 · Cited by 158 · cited 1×
“The native audiovisual navigation task proposed in SoundSpaces 1.0 (Chen et al., 2020) and 2.0 (Chen et al., 2022) requires the agent to reach the location of a continuously ringing phone, as illustrated in Figure 2A.”
MethodInfluentialBengio, Y. (2017) · arXiv · doi:10.48550/arxiv.1709.08568 · Cited by 155 · cited 1×
“…for studying natural intelligence, we believe that now is a prime time to use the latest advances in DL to investigate higher-order thought processes and functions in the brain, such as the access and processing of conscious information (Bengio, 2017; Goyal and Bengio, 2022; Juliani et al., 2022a).”
MethodAUTONOMOUS AGENTS AS EMBODIED AI
Franklin, S. (1997) · Cybernetics & Systems · doi:10.1080/019697297126029 · Cited by 149 · cited 1×
“Embodied agents are agents that have a physical body (real or virtual) with which they interact with their environment (Franklin, 1997).”
BackgroundFacing up to the hard question of consciousness
Dennett, D. C. (2018) · Philosophical Transactions of the Royal Society B: Biological Sciences · doi:10.1098/rstb.2017.0342 · Cited by 149 · cited 1×
“The related “hard question” provides a third possibility of inquiry, seeking explanations for the functional role of consciousness as it manifests in evolved organisms (Dennett, 2018).”
BackgroundDeep Reinforcement Learning at the Edge of the Statistical Precipice
Agarwal, R., Schwarzer, M., Castro, P. S., Courville, A., & Bellemare, M. G. (2021) · arXiv · doi:10.48550/arxiv.2108.13264 · Cited by 142
Deep learning and the global workspace theory
VanRullen, R., & Kanai, R. (2021) · Trends in Neurosciences · doi:10.1016/j.tins.2021.04.005 · Cited by 128
CleanRL: High-quality Single-file Implementations of Deep Reinforcement Learning Algorithms
Huang, S., Dossa, R. F. J., Ye, C., & Braga, J. (2021) · arXiv · doi:10.48550/arxiv.2111.08819 · Cited by 116
Self-citationWhat Is Consciousness, and Could Machines Have It?
Dehaene, S., Lau, H. C., & Kouider, S. (2021) · Springer eBooks · doi:10.1007/978-3-030-54173-6_4 · Cited by 101
Semantic Audio-Visual Navigation
Chen, C., Al-Halah, Z., & Grauman, K. (2021) · 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) · doi:10.1109/cvpr46437.2021.01526 · Cited by 96 · cited 4×
“There are 21 classes of target object categories in total, covering various daily life objects such as chairs, tables, cabinets, sinks, and more (Chen et al., 2021).”
BackgroundMethodInfluentialThe Attention Schema Theory: A Foundation for Engineering Artificial Consciousness
Graziano, M. S. A. (2017) · Frontiers in Robotics and AI · doi:10.3389/frobt.2017.00060 · Cited by 79 · cited 4×
“…overview of various theories of consciousness from a functionalist perspective, including recurrent processing theory (RPT; Lamme, 2006, 2010), GWT (Baars, 1993, 2005), and AST (Graziano, 2017; Wilterson et al., 2020), while also touching upon the aspects of agency and embodiment of such agents.”
BackgroundInfluentialInformation generation as a functional basis of consciousness
Kanai, R., Chang, A., Yu, Y., Magrans de Abril, I., Biehl, M., & Guttenberg, N. (2019) · Neuroscience of Consciousness · doi:10.1093/nc/niz016 · Cited by 72 · cited 1×
“It is within this domain that theorists have proposed models such as GWT (Baars, 1993, 2005), information generation theory (Kanai et al., 2019), and attention schema theory (AST; Graziano, 2017; Wilterson et al., 2020), among others (Rosenthal, 1993; Juliani et al., 2022a; Butlin et al., 2023).”
BackgroundSelf-citationLanguage processing in brains and deep neural networks: computational convergence and its limitsPreprint
Caucheteux, C., & King, J.-R. (2020) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/2020.07.03.186288 · Cited by 71
Blum, L., & Blum, M. (2022) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2115934119 · Cited by 70 · cited 3×
“Similarly, Blum and Blum (2022) emphasized the importance of resource limitations when studying consciousness and related concepts in artificial agents.”
BackgroundMethodMaximum Entropy RL (Provably) Solves Some Robust RL Problems
Eysenbach, B., & Levine, S. (2021) · arXiv · doi:10.48550/arxiv.2103.06257 · Cited by 69
Computational Models of Auditory Scene Analysis: A Review
Szabó, B. T., Denham, S. L., & Winkler, I. (2016) · Frontiers in Neuroscience · doi:10.3389/fnins.2016.00524 · Cited by 65
Coordination among neural modules through a shared global workspace
Goyal, A., Didolkar, A., Lamb, A., Badola, K., Ke, N. R., Rahaman, N., Binas, J., Blundell, C., Mozer, M., & Bengio, Y. (2021) · arXiv · doi:10.48550/arxiv.2103.01197 · Cited by 59
Li, H., & Yong Fan, A. (2019) · NeuroImage · doi:10.1016/j.neuroimage.2019.116059 · Cited by 56
Higher‐order thoughts and the appendage theory of consciousness
ROSENTHAL, D. M. (1993) · Philosophical Psychology · doi:10.1080/09515089308573085 · Cited by 47 · cited 1×
“It is within this domain that theorists have proposed models such as GWT (Baars, 1993, 2005), information generation theory (Kanai et al., 2019), and attention schema theory (AST; Graziano, 2017; Wilterson et al., 2020), among others (Rosenthal, 1993; Juliani et al., 2022a; Butlin et al., 2023).”
BackgroundAttention control and the attention schema theory of consciousness
Wilterson, A. I., Kemper, C. M., Kim, N., Webb, T. W., Reblando, A. M. W., & Graziano, M. S. A. (2020) · Progress in Neurobiology · doi:10.1016/j.pneurobio.2020.101844 · Cited by 42
A convolutional neural-network framework for modelling auditory sensory cells and synapses
Drakopoulos, F., Baby, D., & Verhulst, S. (2021) · Communications Biology · doi:10.1038/s42003-021-02341-5 · Cited by 38
The Perceiver Architecture is a Functional Global Workspace
Cognitive Science Society 2022, & Juliani, A. (2022) · eScholarship (California Digital Library) · doi:10.48448/phj9-7y34 · Cited by 33
Analysing deep reinforcement learning agents trained with domain randomisation
Dai, T., Arulkumaran, K., Gerbert, T., Tukra, S., Behbahani, F., & Bharath, A. A. (2022) · Neurocomputing · doi:10.1016/j.neucom.2022.04.005 · Cited by 31 · cited 1×
“We analyzed how the agent’s representations compare to a standard DL baseline (Alain and Bengio, 2016; Dai et al., 2022; Zhang et al., 2022), as well as its attentional patterns.”
MethodSelf-citationWilterson, A. I., & Graziano, M. S. A. (2021) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.2102421118 · Cited by 26
Neural Mechanisms Underlying Visual Object RecognitionPreprint
Afraz, A., Yamins, D. L. K., & DiCarlo, J. J. (2014) · Cold Spring Harbor Symposia on Quantitative Biology · doi:10.1101/sqb.2014.79.024729 · Cited by 24
On the link between conscious function and general intelligence in humans and machines
Juliani, A., Arulkumaran, K., Sasai, S., & Kanai, R. (2022) · arXiv · doi:10.48550/arxiv.2204.05133 · Cited by 21 · cited 3×
“For example, Transformers (Vaswani et al., 2017) are DL architectures heavily reliant on an attentional mechanism, which is a core component of theories of consciousness such as GWT (Baars, 1993; Juliani et al., 2022a) and AST (Graziano, 2017; Wilterson et al., 2020).”
BackgroundMethodSelf-citationTidal: Tackling Concept Drift in Provenance-Based Advanced Persistent Threats Detection
Zhou, Y., Yu, N., Zhao, T., & Liu, Z. (2026) · DROPS (Schloss Dagstuhl – Leibniz Center for Informatics) · doi:10.4230/oasics.nines.2026.1 · Cited by 18
The Neural MMO Platform for Massively Multiagent ResearchPreprint
Suárez, J., Du, Y., Zhu, C., Mordatch, I., & Isola, P. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2110.07594 · Cited by 14
Monophonic sound source separation with an unsupervised network of spiking neurones
Pichevar, R., & Rouat, J. (2007) · Neurocomputing · doi:10.1016/j.neucom.2007.08.001 · Cited by 13
Explaining V1 Properties with a Biologically Constrained Deep Learning Architecture
Pogoncheff, G., Granley, J., & Beyeler, M. (2023) · neural information processing systems · doi:10.52202/075280-0613 · Cited by 7
Explaining V1 Properties with a Biologically Constrained Deep Learning ArchitecturePreprint
Pogoncheff, G., Granley, J., & Beyeler, M. (2023) · arXiv (Cornell University) · doi:10.48550/arxiv.2305.11275 · Cited by 6
The shift from life in water to life on land advantaged planning in visually-guided behaviorPreprint
Mugan, U., & MacIver, M. A. (2019) · bioRxiv (Cold Spring Harbor Laboratory) · doi:10.1101/585760 · Cited by 5
Lindsay, G. W., Merel, J., Mrsic-Flogel, T., & Sahani, M. (2021) · arXiv · doi:10.48550/arxiv.2112.02027 · Cited by 4
Evaluating Long-Term Memory in 3D Mazes
Pasukonis, J., Lillicrap, T., & Hafner, D. (2022) · arXiv · doi:10.48550/arxiv.2210.13383 · Cited by 3
Comparing human and convolutional neural network performance on scene segmentation
Seijdel, N., Losch, M., de Haan, E. H. F., et al. (2017) · Journal of Vision · doi:10.1167/17.10.1344 · Cited by 2
Light-weight probing of unsupervised representations for reinforcement learning
Zhang, W., GX-Chen, A., Sobal, V., LeCun, Y., & Carion, N. (2022) · arXiv · doi:10.48550/arxiv.2208.12345 · Cited by 2
Bio-inspired Memory Generation by Recurrent Neural Networks
Bedía, M. G., Corchado, J. M., & Castillo, L. (2007) · Springer eBooks · doi:10.1007/978-3-540-73007-1_8 · Cited by 2
What Does BERT Look At? An Analysis of BERT's Attention
Clark, K., Khandelwal, U., Levy, O., et al. (2019) · arXiv · doi:10.48550/arxiv.1906.04341
Research works printed but not linked: 3
- Baker, 2019, Emergent Tool use from Multi-Agent Interaction
- Goodfellow, 2016, Deep Learning
- 2018, OpenAI Five
Cited by
Identifying indicators of consciousness in AI systems
Butlin, P., Long, R., Bayne, T., Bengio, Y., Birch, J., Chalmers, D., Constant, A., Deane, G., Elmoznino, E., Fleming, S. M., et al. (2026) · Trends in Cognitive Sciences · doi:10.1016/j.tics.2025.10.011 · Cited by 56
Self-citationNakanishi, J., Baba, J., Yoshikawa, Y., Kamide, H., & Ishiguro, H. (2025) · Frontiers in Robotics and AI · doi:10.3389/frobt.2025.1607190 · Cited by 3
“…Attention: It is thought that broadcasting allows each specialized module to concentrate its resources on information that is deemed to be extremely important according to the current goals and environmental conditions, thereby improving the efficiency and accuracy of task execution [16], [20].”
A Review of the Current State of Multimodal Data Annotation in Traditional Chinese Medicine
Yutong, D., Yuan, Y., & Hongwei, Z. (2025) · American Journal of Health Research · doi:10.11648/j.ajhr.20251306.17
Goyal, K., Dixit, R. S., Choudhary, S. L., & Goyal, P. (2026) · Journal of Business Analytics · doi:10.1080/2573234x.2026.2617555
Metadata
- Authors
- Dossa, R. F. J., Arulkumaran, K., Juliani, A., Sasai, S., & Kanai, R.
- Affiliations
- Araya Inc. (Japan) (JP)
- Microsoft Research New York City (United States) (US)
- Corresponding author
- Dossa, Rousslan Fernand Julien
- Venue
- Frontiers in Computational Neuroscience
- Publisher
- Frontiers Media SA
- Published
- 2024-06-14
- Volume
- 18
- Pages
- 1352685-1352685
- ISSN
- 1662-5188 (ISSN-L)
- PMC
- PMC11211627
- OpenAlex
- W4399687862
- Cited by
- 5
- References
- 95
- Citations by year
- 2026: 2 · 2025: 3
- FWCI
- 0.73
- Type
- journal article
- Language
- en
- Open access
- gold
- doi.org
- Licence
- Locations
- Frontiers in Computational Neuroscience (publishedVersion, cc-by)
- PubMed Central (publishedVersion, cc-by)
- PubMed (publishedVersion)
- DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Psychiatry, Mental Health, Neuroscience
- Social Sciences › Arts and Humanities › Philosophy › Philosophy and Theoretical Science
- Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
- Keywords
- global workspace theory, consciousness, artificial neural networks, working memory, multimodal environments, recurrent neural networks, attention, representation learning, regularization, task complexity
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
The claim verdicts this report counts are experimental: their accuracy is not yet measured.
A language model drew the claims from the paper on 2026-10-13, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).
A second model labelled each claim on 2026-10-13 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).
Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.