Pith. sign in

REVIEW 2 major objections 6 minor 30 references

When Is a Steerable Concept Representation Real? Measurement Confounds in a Cross-Family Audit of Neuroscience Parallels in LLMs

T0 review · 2 major / 6 minor · reviewed 2026-08-12 · deepseek-v4-flash

Pith's one-line read Apparent 'concept steerability emerges with scale' in LLMs is a measurement artifact: correcting raw units, the readout metric, or the operating point removes the trend, while genuine steering remains significant but trend-free across the…

desk verdict A careful audit that convincingly demonstrates the raw-unit steering scaling law is a measurement artifact; the companion no-trend claim is conditional on unvalidated dose comparability. read the letter →

arxiv 2608.08159 v1 pith:PR2HEX37 submitted 2026-08-08 cs.AI

classification cs.AI
keywords conceptsteeringactivationmeasurementartifactemergentcapabilitieslinearprobingAIneurosciencecross-modelauditresidual-normnormalization
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper sets out to test whether neuroscience-style findings in large language models survive stronger measurement controls, and it argues that the main threat is not missing phenomena but incomparable measurements. In its central case study, an apparent scaling law—concept steerability grows with model size—turns out to depend jointly on raw activation units, the readout metric, and a fixed operating point, and correcting any one of these removes the trend. After normalizing steering strength to the residual-stream norm and selecting the layer and coefficient on held-out concepts, the paper finds concept steering remains significant at every scale tested but shows no significant trend across the dense Qwen3 series, with a slope of +0.31 per doubling and a 95% confidence interval from -0.11 to +0.73. The broader audit finds a linear world map consistently decodable in all 17 checkpoints, number magnitude strongly encoded but with neuron shape depending on selection, and language localization that flips direction under a different attribution method. A sympathetic reader would take away that claims like 'emergence' need calibrated, comparable measurements before they are treated as real.

What carries the argument

The load-bearing device is residual-norm-normalized activation steering: instead of injecting $\alpha d$ in raw units, the paper unit-normalizes the direction and scales the coefficient by the residual-stream norm, so the injection is $h' = h + c\|h\|_\ell \hat{d}$ and its size is exactly a fraction $c$ of the residual norm regardless of model. Around this, the audit protocol adds held-out operating-point selection (choosing layer and strength on one split, scoring on a disjoint split), specificity and directional controls (target direction must raise target over control, the control direction must reverse it, random directions must not), bootstrap confidence intervals, and null comparisons. For the other phenomena, the carrying objects are ridge probes with cross-validated regularization for the world map, a shape-agnostic quadratic-fit criterion plus digit-versus-word cross-format correlation for number tuning, and gradient-by-activation versus activation-magnitude attribution for the lesion study. The mechanism that separates artifact from genuine effect is the comparison of measurements before and after each confound is corrected.

What would settle it

On a single model family trained with a controlled recipe across sizes, rerun the audited steering protocol with residual-norm normalization and held-out operating-point selection; a significant positive slope with a tight confidence interval would overturn the paper's corrected picture. Alternatively, if the raw-unit apparent emergence persists when only the first-token readout is replaced by the specificity contrast, the artifact would not be joint with the readout metric as claimed.

Watch

Extended reading notes

Core claim

The paper's central claim is that the reported finding 'concept steerability emerges with scale' is a measurement artifact rather than a property of the models. Under the conventional recipe—a raw mean-difference direction injected at a raw coefficient into a fixed layer, read out as first-token probability—steerability looks monotone in scale across the Qwen3 ladder, but the injection displaces a fraction of the residual stream that swings non-monotonically between 0.12 and 0.27. Once the intervention is expressed as a fixed fraction of the residual norm, $h' = h + c\|h\|_\ell \hat{d}$, and the (layer, strength) operating point is selected on held-out concepts, the effect is significantly positive at every scale but shows no significant trend across model size (slope +0.31 per doubling, 95% CI $[-0.11, +0.73]$). The paper extends the same audit to three other neuroscience parallels: a linear world map decodes from every checkpoint, number tuning is strong but its bell-versus-monotonic shape depends on the neuron-selection criterion, and language-selective lesional asymmetry reverses under a different attribution method. Its conclusion is that comparability and controls, not new phenomena, are the binding constraint on AI neuroscience.

Load-bearing premise

The load-bearing premise is that an intervention scaled to a fixed fraction of the residual-stream norm delivers a functionally comparable dose across models of different sizes and architectures; the paper itself concedes in its Limitations that this improves comparability but does not guarantee equal functional dose.

Editorial extensions

If this is right

  • Future scaling claims about steerability need calibrated units, held-out operating-point selection, and confidence intervals before the trend is interpretable.
  • Concept steering remains a real, causally effective phenomenon at every tested scale, so the corrected result is not a null result.
  • Pure decoding results, such as the world map, are less vulnerable to audit than intervention-based results, because no selection or operating-point choice enters.
  • The corrected Qwen3 result is an absence of detectable trend, not evidence of scale-invariance, since the confidence interval still admits a moderate positive slope.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Beyond the paper: if the residual-norm dose-comparability assumption holds, the same artifact class should affect other causal representational claims wherever intervention strength is not matched to representation scale.
  • Beyond the paper: a direct test is to run the audited steering protocol with leave-name-out prompts and non-lexical readouts, which the paper lists as construct-validity work, and check whether the flatness across scale survives.
  • Beyond the paper: the four-way taxonomy suggests ranking future 'LLMs also show X' claims by how much unit selection, intervention, and operating-point choice the method involves, since those are the places controls bite.
  • Beyond the paper: the paper's cross-family heterogeneity could partly reflect measurement incomparability rather than genuine family differences; comparing families on matched normalization and matched grids is a natural next step.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

2 major / 6 minor

Summary. The paper audits four neuroscience-inspired interpretability claims across 17 open-weight LLMs from five families, 0.6B to 72B: concept steering, number-magnitude tuning, language localization by lesioning, and the linear world map. The central experiment shows that an apparent scaling law for concept steerability across the Qwen3 ladder is a measurement artifact: with raw activation units, a fixed layer, and a fixed coefficient, steerability appears to grow with scale, but the raw injection displaces a non-monotonic fraction of the residual norm (0.12 to 0.27), and the trend depends jointly on units, readout metric, and operating point. After residual-norm normalization and held-out operating-point selection, steering remains significant at every scale but shows no significant trend across the Qwen3 series (slope +0.31 per doubling, 95% CI [-0.11, +0.73]), with the paper explicitly noting that the null is underpowered. The other results are mixed: number magnitude is strongly encoded but bell-versus-monotonic shape depends on neuron selection; language localization is attribution-dependent; and a linear geographic map is consistently decodable in all 17 models. The paper releases protocol, stimuli, and code.

Significance. The paper makes a valuable methodological contribution to AI neuroscience and interpretability. Its strongest result is the demonstration that a published-style raw-unit steering pipeline manufactures an apparent emergent scaling law, supported by the residual-norm analysis in Fig. 2a, the dose-response and layer-sensitivity curves, held-out operating-point selection, bootstrap confidence intervals, and a bf16-versus-8-bit quantization control. The explicit audit rules R1-R4, the factor decomposition, and the replication across families up to 72B are concrete strengths, as is the candid treatment of nulls and grid-sensitivity. If the artifact claim holds, the paper provides a useful template for comparable cross-model intervention. However, the positive companion claim that no scaling trend exists after correction depends on an untested functional-dose assumption that the paper itself concedes; this limits the force of the headline 'no trend' statement and needs additional work before the claims can be accepted at face value.

major comments (2)
  1. [§3.3 (R2) and §4.1] The 'no detectable trend' conclusion rests on treating the residual-norm-normalized injection h' = h + c||h||_l d_hat as a functionally comparable dose across models, and the paper itself concedes in the Limitations section that residual-norm normalization 'improves comparability but does not guarantee equal functional dose across models.' The held-out operating-point selection in §3.5 and Appendix C mitigates this only partially: it lets each model choose a (layer, strength) cell from a common grid (c in {0.25, 0.5, 1, 2, 4}, with a coarser {0.5, 1, 2} for models at 24B and above), but it does not test whether that grid spans each model's functional dose-response range. If the inverted-U peak in Fig. 2c drifts toward larger c with scale, or if a fixed c||h|| produces a different behavioral change in larger models, the flat slope (+0.31; 95% CI [-0.11, +0.73]) could be an artifact of the corrected pipeline rather than evidence against scale dependence. Please report the selected (layer, c) for each model, check whether selected strengths lie at grid boundaries, expand the c grid for the Qwen3 ladder, and provide a functional-dose sensitivity analysis, for example by calibrating strengths to a common target-token probability lift or by effect-matching across models.
  2. [§4.1 and Fig. 6] The factor decomposition does not isolate which of the three listed confounds (raw units, readout metric, operating point) drives the raw-unit trend, because the readout metric is held fixed at the corrected specificity contrast in all four cells. In particular, the raw+fixed cell in Fig. 6 already uses the specificity contrast and is flat, so the decomposition cannot support the statement that the apparent scaling is 'not attributable to any single factor.' A proper decomposition would vary one factor at a time from the exact naive pipeline (raw units, first-token readout, fixed layer and strength) and report the slope for each single correction. This is important because the abstract and Section 4.1 claim that correcting any one of the three choices removes the trend.
minor comments (6)
  1. [§2] The 'activation addition' citation appears as a literal '(?)' placeholder in the Related Work section; please supply the intended reference.
  2. [§3.6] The sentence describing the memory-efficient lesion attribution is garbled: 'requires grad' should be 'requires_grad', and the phrase 'requires grad removes that buffer' should be reworded for clarity.
  3. [Figures 11 and 12 (Appendix D)] The token-by-token examples in Figures 11 and 12 contain unrendered glyph or token-id sequences (e.g., '/uni00000011') rather than readable text; please fix the rendering so the qualitative examples are legible.
  4. [Table 1] The pass-rate column mixes fine-grid and coarse-grid values, and the table notes that absolute pass-rates are not comparable across these groups; consider adding a visual separator or repeating the caveat directly in the column header so it is not missed by readers.
  5. [§4.1 and Abstract] The 'no significant trend' wording is used in several summary locations without always carrying the power caveat. Since the 95% CI does not exclude +0.73 per doubling and Appendix C notes that 3-4x more concepts or models would be needed to exclude a +0.3 trend, please consistently phrase the result as 'no detectable trend, with a confidence interval that admits a moderate positive slope' in the abstract, Table 2, and the conclusion.
  6. [Table 1] The world-map R2 column lists values such as '0.53/0.67' without stating in the caption that the two numbers are latitude and longitude; please make that explicit.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the artifact claim rests on an independent calibration argument and the no-trend result is an empirical held-out regression, not a fitted prediction.

full rationale

The paper's central artifact claim is supported by an independent calibration argument: Section 4.1 shows the raw injection fraction ||d||/||h|| varies non-monotonically across the Qwen3 ladder between 0.12 and 0.27, so a fixed raw coefficient is not a comparable dose, and the factor decomposition (Fig. 6) evaluates four combinations of units and operating point with none showing a significant scaling slope. The corrected no-trend result is an empirical regression on per-model held-out effects, not a fitted parameter renamed as a prediction; operating points are selected on held-out concepts and scored on disjoint folds. The residual-norm normalization (R2, Section 3.4, h' = h + c||h|| d_hat) is a stated methodological choice, not derived from the trend outcome, and the paper explicitly concedes in Limitations that 'residual-norm normalization improves comparability but does not guarantee equal functional dose across models,' which is a construct-validity caveat rather than a circular step. The efficient-coding derivation (Appendix A) is presented as a normative model with stated assumptions, not fitted to the data. References are to external prior work (Gurnee and Tegmark, Rimsky et al., Schaeffer et al.); no load-bearing self-citation chain or uniqueness theorem is invoked. No equation in the paper reduces to its inputs by construction, so no circularity is found; the skeptical concern about functional dose is a substantive empirical limitation that belongs in correctness risk, not circularity.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The paper introduces no new physical or theoretical entities: its contributions are measurement controls (residual-norm normalization, held-out operating-point selection) and an audit taxonomy. The main free parameters are the operating-point hyperparameters and shape-classification thresholds, all disclosed as grid-selected or threshold-based. The axioms are domain assumptions inherited from the probing and steering literature, plus the normalization-comparability assumption the paper itself flags as imperfect.

free parameters (4)
  • steering strength c (residual-norm fraction) = selected per model on held-out concepts from grid {0.25,0.5,1,2,4}
    The intervention magnitude is expressed as c times the residual norm; c is chosen by held-out operating-point selection, so it is a tuned operating point rather than a fitted constant.
  • intervention layer fraction = approximately 0.8 depth for all Qwen3 models
    The layer at which the steering direction is injected is selected on held-out concepts; the paper shows a fixed layer (2/3) is not comparable across models.
  • bell-neuron quadratic-gain threshold = 0.15 R^2 gain over linear fit
    A quadratic fit must beat a linear fit by more than 0.15 to classify a neuron as bell-shaped; the paper says the exact fraction is threshold-sensitive while the existence of bell units is not.
  • cross-format invariance cutoff = r > 0.5
    A digit-vs-word tuning correlation above 0.5 is used to label bell units as partly format-invariant; the paper calls this an arbitrary cutoff and reports sensitivity qualitatively.
assumptions (5)
  • domain assumption Residual-norm-normalized interventions are functionally comparable across models of different sizes and architectures.
    Used throughout Section 3.3 and 4.1 to justify removing the raw-unit scaling; the paper explicitly flags in Limitations that this normalization does not guarantee equal functional dose.
  • domain assumption The Qwen3 ladder checkpoints form a usable size axis despite being independently trained runs.
    The trend regression treats log2 model size as the independent variable across Qwen3 checkpoints; the paper states these are separate training runs and not a controlled size sweep.
  • domain assumption The linear representation hypothesis (Park et al.) holds well enough that adding a direction to the residual stream is a causal concept intervention.
    All steering experiments assume that a population direction d, built as a mean difference, corresponds to a manipulable concept; this is the standard basis of the steering literature.
  • domain assumption Gradient x activation attribution measures the causal importance of a neuron for a language.
    The language lesion experiments rank neurons by this attribution and then ablate them; the paper shows that an alternative attribution reverses the asymmetry, making this assumption load-bearing.
  • domain assumption The efficient-coding model in Appendix A assumes a scalar readout coordinate, additive readout noise, and reward proportional to number frequency.
    This derivation predicts a monotonic log-compressed code; the paper labels it normative and not a description of the models, so it is a stated idealized model rather than a hidden assumption.

how reviews work

0 comments
Cite this review

Pith. "Pith review of When Is a Steerable Concept Representation Real? Measurement Confounds in a Cross-Family Audit of Neuroscience Parallels in LLMs." pith.science (2026). https://pith.science/paper/PR2HEX37

@misc{pith2026260808159,
  author       = {Pith},
  title        = {Pith review of: When Is a Steerable Concept Representation Real? Measurement Confounds in a Cross-Family Audit of Neuroscience Parallels in LLMs},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/PR2HEX37}},
  note         = {Machine review of arXiv:2608.08159}
}
abstract

Large language models (LLMs) are increasingly reported to exhibit human-like neural and cognitive signatures, including concept cells, mental number lines, and cognitive maps. These claims often rely on linear probing and activation steering applied to a single model, yet both methods are highly sensitive to measurement choices. A reported parallel may therefore reflect the model, the measurement procedure, or both. We audit four representative neuroscience-inspired paradigms across 17 models from five families, spanning $0.6$B to $72$B parameters. Our main experiment examines the causal steerability of concept directions. With raw activation units and a fixed layer and coefficient, steerability appears to increase with model scale, resembling an emergent capability. However, this pattern is produced by an uncalibrated pipeline rather than by a claim established in the steering literature. The trend depends jointly on raw units, the readout metric, and the operating point; correcting any one of these removes it. With residual-norm-comparable interventions and held-out operating-point selection, concept steering remains significant at every scale, but shows no significant trend across the Qwen3 series, although the confidence interval does not rule out a moderate positive slope. The remaining results are mixed. A linear geographic world map is consistently decodable in every tested checkpoint up to $72$B. Number magnitude is strongly encoded, but whether individual neurons appear bell-shaped or monotonic depends on the selection criterion. Language-specific structure is localizable, but the direction of the cross-lingual asymmetry reverses under a different attribution method. These results suggest that the main constraint on AI neuroscience is not a lack of phenomena, but a lack of comparable measurements and adequate controls. We release the protocol, stimuli, and code.

Figures

Figures reproduced from arXiv: 2608.08159 by the authors.

Figure 1
Figure 1. Same data, different ruler, different conclusion. Left: a standard pipeline makes concept steerability appear to increase with model size in raw activation units. Once the intervention is normalized across models and the operating point is selected on held-out concepts, no significant scaling trend remains. Right: the four neuroscience-inspired phenomena examined in this study and the conclusions supported after con… view at source ↗
Figure 2
Figure 2. Main experiment: the “concept steerability emerges with scale” effect is a measurement artifact. (a) The units confound: the raw injection fraction ∥d∥/∥h∥ varies non-monotonically across the Qwen3 ladder, so a fixed raw coefficient is an uncontrolled intervention. (b) In raw units the effect nonetheless appears to grow monotonically with scale, an apparent emergent capability. (c) Under residual-norm normalization … view at source ↗
Figure 3
Figure 3. Magnitude: strong tuning, mixed shape. (a) Held-out magnitude tuning is strong in every model, to 72B. (b) Under shape-agnostic neuron selection, bell-shaped number neurons appear, common in Lla￾ma/Phi/Mistral and rare in Qwen; the linear-correlation selection that returned “0% bell” was biased. models (CI straddles 0; Fig. 4a) and significantly positive in only five (Qwen3-0.6B, Ministral-8B, Qwen3-32B, Qwen2.5-72B… view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Language lesioning (gradient×activation): only Chinese is reliably localizable, and this asym￾metry itself flips under a different attribution (App. C). (a) English lesion selectivity with bootstrap 95% CIs (44 sentences/language); filled = CI excludes 0. It is null in…
Figure 6
Figure 6. Figure 6: Factor decomposition: which measure￾ment choice manufactures the apparent scaling? Holding the metric (specificity contrast) and concept set (24 concepts) constant, we vary two factors: unit normalization (raw vs. ∥h∥-normalized) and operating￾point selection (fixed L·…
Figure 7
Figure 7. Figure 7: Cross-family audit at scale (0.6–72B), from the large-model controls. (a) Held-out steering effect stays significantly positive at every scale (filled = fine grid ≤ 14B; open = coarser grid ≥ 24B). (b) Steering quantization control: Qwen3-32B in 8-bit vs. bf16 is nearl…
Figure 9
Figure 9. Figure 9: Steering effect size, not just pass-rate. Per￾concept held-out specificity effect (24 concepts; bar = mean) across the Qwen3 ladder; the distributions over￾lap across scale. not drive the steering result. Because the grid is coarse and mixes families, we read significa…
Figure 10
Figure 10. Figure 10: What steering does, in words (real Llama-3.1-8B generations, one fixed prompt). A concept direction added to the residual stream injects that concept’s vocabulary (highlighted) while the sentence stays coherent at a moderate coefficient. The last row shows over-steeri…
Figure 11
Figure 11. Figure 11: Real number-neuron tuning curves (Llama-3.1-8B). A bell-shaped and a monotonic unit; each neuron’s digit and word tuning curves nearly co￾incide (r = 0.96), indicating magnitude tuning beyond token identity. les west, Tokyo and Beijing east, Moscow north, Cape Town so…
Figure 13
Figure 13. Figure 13: The recovered world map (Qwen3-1.7B, a representative small model). Cross-validated predicted coordinates of held-out cities; colour encodes true lon￾gitude. Even at 1.7B, geography is recovered (R2 = 0.62/0.66 for latitude and longitude). note 1). We evaluate 17 open…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

30 extracted references · 8 canonical work pages

  1. [1]

    Nature , volume=

    Invariant visual representation by single neurons in the human brain , author=. Nature , volume=. 2005 , publisher=

  2. [2]

    , author=

    The hippocampus as a spatial map: preliminary evidence from unit activity in the freely-moving rat. , author=. Brain research , year=

  3. [3]

    Nature , volume=

    Microstructure of a spatial map in the entorhinal cortex , author=. Nature , volume=. 2005 , publisher=

  4. [4]

    Science , volume=

    Organizing conceptual knowledge in humans with a gridlike code , author=. Science , volume=. 2016 , publisher=

  5. [5]

    Cell , volume=

    The geometry of abstraction in the hippocampus and prefrontal cortex , author=. Cell , volume=. 2020 , publisher=

  6. [6]

    A simple coding procedure enhances a neuron’s information capacity , author=. Z. Naturforsch , volume=

  7. [7]

    Trends in cognitive sciences , volume=

    The neural basis of the Weber--Fechner law: a logarithmic mental number line , author=. Trends in cognitive sciences , volume=. 2003 , publisher=

  8. [8]

    Journal of neuroscience , volume=

    Neuronal avalanches in neocortical circuits , author=. Journal of neuroscience , volume=. 2003 , publisher=

Show all 30 references
  1. [9]

    Nature , volume=

    The capacity of visual working memory for features and conjunctions , author=. Nature , volume=. 1997 , publisher=

  2. [10]

    International Conference on Learning Representations , volume=

    Language models represent space and time , author=. International Conference on Learning Representations , volume=

  3. [11]

    Proceedings of the 41st International Conference on Machine Learning , articleno =

    Park, Kiho and Choe, Yo Joong and Veitch, Victor , title =. Proceedings of the 41st International Conference on Machine Learning , articleno =. 2024 , publisher =

  4. [12]

    Nature Reviews Neuroscience , volume=

    The neuronal code for number , author=. Nature Reviews Neuroscience , volume=. 2016 , publisher=

  5. [13]

    Transactions on Machine Learning Research , year=

    Emergent Abilities of Large Language Models , author=. Transactions on Machine Learning Research , year=

  6. [14]

    Advances in neural information processing systems , volume=

    Are emergent abilities of large language models a mirage? , author=. Advances in neural information processing systems , volume=

  7. [15]

    arXiv preprint arXiv:2507.11771 , year=

    Scaling laws for activation steering with Llama 2 models and refusal mechanisms , author=. arXiv preprint arXiv:2507.11771 , year=

  8. [16]

    Activation addition: Steering language models without optimization , author=

  9. [17]

    arXiv preprint arXiv:2310.01405 , year=

    Representation engineering: A top-down approach to ai transparency , author=. arXiv preprint arXiv:2310.01405 , year=

  10. [18]

    Advances in Neural Information Processing Systems , volume=

    Inference-time intervention: Eliciting truthful answers from a language model , author=. Advances in Neural Information Processing Systems , volume=

  11. [19]

    International Conference on Learning Representations (ICLR) Workshop , year =

    Understanding Intermediate Layers Using Linear Classifier Probes , author =. International Conference on Learning Representations (ICLR) Workshop , year =

  12. [20]

    Designing and interpreting probes with control tasks , author=. Proceedings of the 2019 conference on empirical methods in natural language processing and the 9th international joint conference on natural language processing (emnlp-ijcnlp) , pages=

  13. [21]

    The Eleventh International Conference on Learning Representations , year=

    Emergent World Representations: Exploring a Sequence Model Trained on a Synthetic Task , author=. The Eleventh International Conference on Learning Representations , year=

  14. [22]

    Transformer Circuits Thread , year=

    Towards Monosemanticity: Decomposing Language Models with Dictionary Learning , author=. Transformer Circuits Thread , year=

  15. [23]

    arXiv preprint arXiv:2502.16147 , year=

    Number representations in llms: A computational parallel to human perception , author=. arXiv preprint arXiv:2502.16147 , year=

  16. [24]

    arXiv preprint arXiv:2310.06824 , year=

    The geometry of truth: Emergent linear structure in large language model representations of true/false datasets , author=. arXiv preprint arXiv:2310.06824 , year=

  17. [25]

    Advances in neural information processing systems , volume=

    Locating and editing factual associations in gpt , author=. Advances in neural information processing systems , volume=

  18. [26]

    Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

    Steering llama 2 via contrastive activation addition , author=. Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages=

  19. [27]

    Transformer Circuits Thread , year=

    Toy Models of Superposition , author=. Transformer Circuits Thread , year=

  20. [28]

    Scaling Monosemanticity: Extracting Interpretable Features from

    Templeton, Adly and Conerly, Tom and Marcus, Jonathan and others , journal=. Scaling Monosemanticity: Extracting Interpretable Features from

  21. [29]

    Transformer Circuits Thread , year=

    Emotion Concepts and their Function in a Large Language Model , author=. Transformer Circuits Thread , year=

  22. [30]

    Forty-first International Conference on Machine Learning , year=

    Position: The platonic representation hypothesis , author=. Forty-first International Conference on Machine Learning , year=

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.