Improve flow of AFFECTOR purpose and research context

SHA256SUMS.txt

@@ -1,7 +1,7 @@
 321cd041f8060331e034b2a72578f57cd3e498883e660016cf73f0a8ba69a9a6  README.md
 e059179343c7acd54f1d6b8da8ba36cdb0fe941aca291f9be6d9bf4aa5a9dd0b  docs/AFFECTOR-Technical-Paper.pdf
-75efa70898e117dac39373c4fe3eba88e79fcee0f2cf22d55762794206450d7a  docs/TECHNICAL_PAPER.md
-75efa70898e117dac39373c4fe3eba88e79fcee0f2cf22d55762794206450d7a  docs/TECHNICAL_REPORT.md
+1efa350b07530950ea404fe8bf40c54550c99856a15cc07a2d9829bc0d4c7cdd  docs/TECHNICAL_PAPER.md
+1efa350b07530950ea404fe8bf40c54550c99856a15cc07a2d9829bc0d4c7cdd  docs/TECHNICAL_REPORT.md
 ed8fa321af4fc4e4703efcc22037e244fd072c936d11afc566b8eb862f3f0b1b  downloads/AFFECTOR-macOS-AppleSilicon.zip
 d7621338f71fdc31ce9b6e1d53f15d61d6a76344f2111520fb62f9b00293853b  downloads/AFFECTOR-source.zip
 585966103f53f58a9983bba63e3d27dcb2020033a2f64c6dc63684afe418ae14  paper/AFFECTOR-ACM-source.zip
@@ -18,7 +18,7 @@ e2c0590754823491af4657a764ab6682e27c5eaf0173fc9f5f03a62e84ccb308  paper/ILLUSTRA
 d14edda139a36c0bac8b56d02e5cda63ab7558c6db1424f1087189a9429bfaf0  paper/LITERATURE_SOURCES.md
 578a471b301cdf67305c2b997fce39494c2fd0d15eaf56836379943b3d10f90c  paper/PAPER-SHA256SUMS.txt
 22f8f7bf6992778eb1190ee02b6e83dacbd54ff34af7777a18b4e27a195309a3  paper/README.md
-27090c2045d62c461ab3c34410bda3d0fa7c4fc7d5656af720cd6fba5d3e99d2  paper/TECHNICAL_PAPER.md
+97a2a8a88f99be148a9b8a5774cd276cc38317712132ba024bd0b73b88a05760  paper/TECHNICAL_PAPER.md
 4bd1a1b7f6af31b12b9c92bcd978fe116482892b3463f8ec9c922bc1f636c7a0  paper/figure-specifications.json
 b11798a28d94d60110a7b19f58107d5e60926b68265f64e62f59e9230a8f500d  paper/figures/illustration-01.pdf
 d2ae42a17ff3f16b89a7cb85d21ac1f3c7a19c2bc62d61f1cc581c2ca6d3e596  paper/figures/illustration-01.png
@@ -32,4 +32,4 @@ a6a497a38f4ba9de8088f410535bee13dd862d1a97f4175d205df768e1d2c1c9  paper/figures/
 7d8f30b829993326e8d764ae18b3974acddbd86b760d4071bdff4edfcb9b299a  paper/figures/illustration-05.png
 9257f38423bd7e496f1a0916c0fbaefbfb32ab544a684ca9a411508201f0535e  paper/references.bib
 eed8b979a5a299b42e925d9027332e7df876400dcef046f2cfe26beef91d1189  paper/render_figures.py
-27090c2045d62c461ab3c34410bda3d0fa7c4fc7d5656af720cd6fba5d3e99d2  paper/source.md
+97a2a8a88f99be148a9b8a5774cd276cc38317712132ba024bd0b73b88a05760  paper/source.md

docs/AFFECTOR.pdf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8f7bbc77f3fc600efd5075b9a9230e29cae4f46d0b850721a0f26dc9651c977
-size 1948134
+oid sha256:632798acebaa3cbc687514ac8bdc029dd5dfd6a871bbb3ae7b9f323f5f6d616e
+size 1948806

docs/TECHNICAL_PAPER.md

@@ -5,7 +5,7 @@
 **William Taggart**  
 Modern and Classical Languages and Literatures, Virginia Tech  
 Email: wtaggart@vt.edu  
-November 13, 2026 · Implementation report · Desktop version 0.1.0
+November 14, 2026 · Implementation report · Desktop version 0.1.0
 
 ## Abstract
 
@@ -13,21 +13,21 @@ AFFECTOR investigates whether a prescribed emotional trajectory can influence th
 
 ## Purpose and Research Context
 
-A document conveys more than propositions: lexical choice, rhythm, imagery, focus, and narrative stance contribute to its emotional character. AFFECTOR investigates whether a user-defined emotional progression can guide a rewrite while leaving room to choose how much of the original content to preserve. It represents that progression as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. Such a curve specifies a coarse target rather than a unique text, so emotional fit must be considered alongside readability, content change, and output length.
+A document conveys more than propositions. Its lexical choices, rhythm, imagery, focus, and narrative stance also shape how a reader feels as the text unfolds. AFFECTOR begins with a practical question about that progression: can a writer specify a desired emotional path and use it to guide a rewrite, while choosing how much of the original content to preserve? To make this question tractable, the application represents the path as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. This representation supplies a coarse target for experimentation, leaving the model to realize it in language. The resulting text must therefore be judged both by its emotional fit and by its readability, content changes, and length.
 
-Activation steering provides one mechanism for pursuing this target. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). AFFECTOR uses contrastive calibration to map emotional coordinates into changes to a frozen model's internal states during local inference. Its API backend instead describes the coordinates in instructions to a hosted model. These two routes support comparison between numerical intervention and prompt-based control, but their strengths and outcomes cannot be treated as equivalent.
+Activation steering offers a way to connect this target to the process of generation. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). Building on this contrastive approach, AFFECTOR maps emotional coordinates into changes to a frozen model's internal states during local inference. The application also provides an API route, in which the coordinates are expressed as instructions to a hosted model. Together, these routes make it possible to investigate numerical intervention alongside prompt-based control, although their settings and outcomes are not directly interchangeable.
 
 ![Document workspace with an example input, rewrite controls, and an empty output pane before generation.](../paper/figures/screenshot-01.png)
 
 *Document workspace with an example input, rewrite controls, and an empty output pane before generation.*
 
-The choice to expose emotional coordinates is informed by work connecting model representations with human judgments. Wu and colleagues relate model emotion spaces to human valence/arousal ratings and report concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues assess emotional intensity and text quality with human raters, finding effects that depend on the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These studies motivate separate assessments of emotional movement and readable expression. They do not validate AFFECTOR's inward/outward axis or establish a correspondence between their intervention scales and its RMS-scaled controls.
+For either route, the usefulness of the coordinates depends on their relationship to human interpretations of the output. Work by Wu and colleagues connects model emotion spaces with human valence/arousal ratings and reports concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues extend this evaluative perspective by examining emotional intensity and text quality with human raters, finding effects that vary with the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These findings motivate assessing emotional movement and readable expression separately. They also clarify what remains to be established for AFFECTOR: neither its inward/outward axis nor the correspondence between its RMS-scaled controls and a reader's emotional judgment has been validated by those studies.
 
-The practical need for bounded generation follows from evidence that emotional intervention can change more than tone. E-STEER studies effects on reasoning, generation, safety, and agent behavior and reports non-monotonic relationships ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). AFFECTOR therefore treats repetition, truncation, and failed generations as outcomes to record, rather than assuming that stronger intervention produces a closer emotional match. Its rewrite styles make the competing requirements of emotional transformation and content preservation explicit.
+This uncertainty concerns the intervention itself as well as its evaluation. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition, while Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3); [Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Thus, recognizing an emotion in a representation does not by itself determine where to intervene to change it. The partially coupled supportive-empathy directions reported by Ha and colleagues further raise the question of whether changing one emotional dimension also moves another ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)). Research on automated vector extraction and context-dependent strength offers ways to investigate these problems, with benefits that depend on the task and intended trade-off ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3); [Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). AFFECTOR exposes layer choices and coordinate mixtures as experimental controls; testing their interactions and potential adaptation remains part of the research agenda.
 
-Choosing an evaluator and choosing an intervention site are also separate research problems. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3)). Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Meanwhile, Ha and colleagues find partially coupled supportive-empathy directions ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)), Cui and Chen automate vector extraction with task-dependent benefits ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3)), and Ferrando and colleagues study context-dependent strength to balance toxicity mitigation and utility ([Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). Together, this work motivates testing layer choice, cross-axis interference, and adaptation. The current coordinate mixture is a simpler experimental control, rather than an implementation of those learned selectors or gates.
+When these interventions are applied throughout a document, their effects must also be considered over the course of generation. E-STEER reports non-monotonic relationships between emotional interventions and outcomes in reasoning, generation, safety, and agent behavior ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Similarly, Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on the architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). These results explain why greater intervention strength cannot simply be equated with a closer emotional match. In response to this practical difficulty, AFFECTOR combines bounded generation with records of repetition, truncation, and failed attempts. Its rewrite styles also let the user state how much content preservation should constrain the transformation, making that trade-off explicit when reviewing an output.
 
-The report consequently focuses on the implemented transformation pipeline and its observable limits. The application does not automatically score each generated passage and iteratively minimize its emotional distance; verification is a separate user-invoked operation. SteerScope's joint evaluation of efficacy, side effects, generalization, and data dependence supplies an appropriate broader evaluation agenda ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs reinforces the need for prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Those comparisons remain research requirements, not completed evidence of AFFECTOR's effectiveness.
+Taken together, these considerations define the scope of the present report: the implemented transformation pipeline, the choices it exposes, and the limits visible in its operation. Emotional verification is a separate user-invoked step; the application does not automatically score each passage and iteratively minimize its distance from the target. A broader assessment must therefore compare the requested progression with independently judged outcomes. SteerScope provides a useful framework by evaluating efficacy alongside side effects, generalization, and dependence on steering data ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs supports including prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Such comparisons would establish whether AFFECTOR's controls offer a reliable advantage for particular writing tasks; they remain research requirements rather than completed evidence of effectiveness.
 
 ## Representation and alignment
 

docs/TECHNICAL_REPORT.md

@@ -5,7 +5,7 @@
 **William Taggart**  
 Modern and Classical Languages and Literatures, Virginia Tech  
 Email: wtaggart@vt.edu  
-November 13, 2026 · Implementation report · Desktop version 0.1.0
+November 14, 2026 · Implementation report · Desktop version 0.1.0
 
 ## Abstract
 
@@ -13,21 +13,21 @@ AFFECTOR investigates whether a prescribed emotional trajectory can influence th
 
 ## Purpose and Research Context
 
-A document conveys more than propositions: lexical choice, rhythm, imagery, focus, and narrative stance contribute to its emotional character. AFFECTOR investigates whether a user-defined emotional progression can guide a rewrite while leaving room to choose how much of the original content to preserve. It represents that progression as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. Such a curve specifies a coarse target rather than a unique text, so emotional fit must be considered alongside readability, content change, and output length.
+A document conveys more than propositions. Its lexical choices, rhythm, imagery, focus, and narrative stance also shape how a reader feels as the text unfolds. AFFECTOR begins with a practical question about that progression: can a writer specify a desired emotional path and use it to guide a rewrite, while choosing how much of the original content to preserve? To make this question tractable, the application represents the path as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. This representation supplies a coarse target for experimentation, leaving the model to realize it in language. The resulting text must therefore be judged both by its emotional fit and by its readability, content changes, and length.
 
-Activation steering provides one mechanism for pursuing this target. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). AFFECTOR uses contrastive calibration to map emotional coordinates into changes to a frozen model's internal states during local inference. Its API backend instead describes the coordinates in instructions to a hosted model. These two routes support comparison between numerical intervention and prompt-based control, but their strengths and outcomes cannot be treated as equivalent.
+Activation steering offers a way to connect this target to the process of generation. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). Building on this contrastive approach, AFFECTOR maps emotional coordinates into changes to a frozen model's internal states during local inference. The application also provides an API route, in which the coordinates are expressed as instructions to a hosted model. Together, these routes make it possible to investigate numerical intervention alongside prompt-based control, although their settings and outcomes are not directly interchangeable.
 
 ![Document workspace with an example input, rewrite controls, and an empty output pane before generation.](../paper/figures/screenshot-01.png)
 
 *Document workspace with an example input, rewrite controls, and an empty output pane before generation.*
 
-The choice to expose emotional coordinates is informed by work connecting model representations with human judgments. Wu and colleagues relate model emotion spaces to human valence/arousal ratings and report concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues assess emotional intensity and text quality with human raters, finding effects that depend on the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These studies motivate separate assessments of emotional movement and readable expression. They do not validate AFFECTOR's inward/outward axis or establish a correspondence between their intervention scales and its RMS-scaled controls.
+For either route, the usefulness of the coordinates depends on their relationship to human interpretations of the output. Work by Wu and colleagues connects model emotion spaces with human valence/arousal ratings and reports concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues extend this evaluative perspective by examining emotional intensity and text quality with human raters, finding effects that vary with the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These findings motivate assessing emotional movement and readable expression separately. They also clarify what remains to be established for AFFECTOR: neither its inward/outward axis nor the correspondence between its RMS-scaled controls and a reader's emotional judgment has been validated by those studies.
 
-The practical need for bounded generation follows from evidence that emotional intervention can change more than tone. E-STEER studies effects on reasoning, generation, safety, and agent behavior and reports non-monotonic relationships ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). AFFECTOR therefore treats repetition, truncation, and failed generations as outcomes to record, rather than assuming that stronger intervention produces a closer emotional match. Its rewrite styles make the competing requirements of emotional transformation and content preservation explicit.
+This uncertainty concerns the intervention itself as well as its evaluation. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition, while Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3); [Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Thus, recognizing an emotion in a representation does not by itself determine where to intervene to change it. The partially coupled supportive-empathy directions reported by Ha and colleagues further raise the question of whether changing one emotional dimension also moves another ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)). Research on automated vector extraction and context-dependent strength offers ways to investigate these problems, with benefits that depend on the task and intended trade-off ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3); [Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). AFFECTOR exposes layer choices and coordinate mixtures as experimental controls; testing their interactions and potential adaptation remains part of the research agenda.
 
-Choosing an evaluator and choosing an intervention site are also separate research problems. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3)). Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Meanwhile, Ha and colleagues find partially coupled supportive-empathy directions ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)), Cui and Chen automate vector extraction with task-dependent benefits ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3)), and Ferrando and colleagues study context-dependent strength to balance toxicity mitigation and utility ([Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). Together, this work motivates testing layer choice, cross-axis interference, and adaptation. The current coordinate mixture is a simpler experimental control, rather than an implementation of those learned selectors or gates.
+When these interventions are applied throughout a document, their effects must also be considered over the course of generation. E-STEER reports non-monotonic relationships between emotional interventions and outcomes in reasoning, generation, safety, and agent behavior ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Similarly, Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on the architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). These results explain why greater intervention strength cannot simply be equated with a closer emotional match. In response to this practical difficulty, AFFECTOR combines bounded generation with records of repetition, truncation, and failed attempts. Its rewrite styles also let the user state how much content preservation should constrain the transformation, making that trade-off explicit when reviewing an output.
 
-The report consequently focuses on the implemented transformation pipeline and its observable limits. The application does not automatically score each generated passage and iteratively minimize its emotional distance; verification is a separate user-invoked operation. SteerScope's joint evaluation of efficacy, side effects, generalization, and data dependence supplies an appropriate broader evaluation agenda ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs reinforces the need for prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Those comparisons remain research requirements, not completed evidence of AFFECTOR's effectiveness.
+Taken together, these considerations define the scope of the present report: the implemented transformation pipeline, the choices it exposes, and the limits visible in its operation. Emotional verification is a separate user-invoked step; the application does not automatically score each passage and iteratively minimize its distance from the target. A broader assessment must therefore compare the requested progression with independently judged outcomes. SteerScope provides a useful framework by evaluating efficacy alongside side effects, generalization, and dependence on steering data ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs supports including prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Such comparisons would establish whether AFFECTOR's controls offer a reliable advantage for particular writing tasks; they remain research requirements rather than completed evidence of effectiveness.
 
 ## Representation and alignment
 

paper/AFFECTOR-source.zip

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f224f098413cc47d60d52dd48283b595f815574cb709c2357385d930a38bc701
-size 4205686
+oid sha256:7a96304eae89114abc39be017b5163b88388efbfbb0ef017fae59ecfaa0721c6
+size 4207478

paper/AFFECTOR.pdf

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:c8f7bbc77f3fc600efd5075b9a9230e29cae4f46d0b850721a0f26dc9651c977
-size 1948134
+oid sha256:632798acebaa3cbc687514ac8bdc029dd5dfd6a871bbb3ae7b9f323f5f6d616e
+size 1948806

paper/AFFECTOR.tex

@@ -35,7 +35,7 @@ Activation Interventions}
 \affiliation{\department{Modern and Classical Languages and Literatures}
 \institution{Virginia Tech}\city{Blacksburg}\state{Virginia}\country{USA}}
 \email{wtaggart@vt.edu}
-\subtitle{Technical Report: November 13, 2026}
+\subtitle{Technical Report: November 14, 2026}
 \begin{abstract}
 AFFECTOR investigates whether a prescribed emotional trajectory can
 influence the rewriting of a document. A trajectory assigns valence,
@@ -56,9 +56,9 @@ or controller of human emotion.
 \maketitle
 \section{Purpose and Research Context}\label{literature-review-activation-steering-and-emotion}\label{problem-formulation}
 
-A document conveys more than propositions: lexical choice, rhythm, imagery, focus, and narrative stance contribute to its emotional character. AFFECTOR investigates whether a user-defined emotional progression can guide a rewrite while leaving room to choose how much of the original content to preserve. It represents that progression as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. Such a curve specifies a coarse target rather than a unique text, so emotional fit must be considered alongside readability, content change, and output length.
+A document conveys more than propositions. Its lexical choices, rhythm, imagery, focus, and narrative stance also shape how a reader feels as the text unfolds. AFFECTOR begins with a practical question about that progression: can a writer specify a desired emotional path and use it to guide a rewrite, while choosing how much of the original content to preserve? To make this question tractable, the application represents the path as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. This representation supplies a coarse target for experimentation, leaving the model to realize it in language. The resulting text must therefore be judged both by its emotional fit and by its readability, content changes, and length.
 
-Activation steering provides one mechanism for pursuing this target. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples \cite{turner2023actadd,panickssery2024caa}. AFFECTOR uses contrastive calibration to map emotional coordinates into changes to a frozen model's internal states during local inference. Its API backend instead describes the coordinates in instructions to a hosted model. These two routes support comparison between numerical intervention and prompt-based control, but their strengths and outcomes cannot be treated as equivalent.
+Activation steering offers a way to connect this target to the process of generation. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples \cite{turner2023actadd,panickssery2024caa}. Building on this contrastive approach, AFFECTOR maps emotional coordinates into changes to a frozen model's internal states during local inference. The application also provides an API route, in which the coordinates are expressed as instructions to a hosted model. Together, these routes make it possible to investigate numerical intervention alongside prompt-based control, although their settings and outcomes are not directly interchangeable.
 
 \begin{figure}[!htb]
 \centering
@@ -68,13 +68,13 @@ Activation steering provides one mechanism for pursuing this target. Turner and
 \label{fig:app-screen-1}
 \end{figure}
 
-The choice to expose emotional coordinates is informed by work connecting model representations with human judgments. Wu and colleagues relate model emotion spaces to human valence/arousal ratings and report concept-based steering across linguistic-cultural settings \cite{wu2025emotion}. Diallo and colleagues assess emotional intensity and text quality with human raters, finding effects that depend on the emotion and steering strength \cite{diallo2026style}. These studies motivate separate assessments of emotional movement and readable expression. They do not validate AFFECTOR's inward/outward axis or establish a correspondence between their intervention scales and its RMS-scaled controls.
+For either route, the usefulness of the coordinates depends on their relationship to human interpretations of the output. Work by Wu and colleagues connects model emotion spaces with human valence/arousal ratings and reports concept-based steering across linguistic-cultural settings \cite{wu2025emotion}. Diallo and colleagues extend this evaluative perspective by examining emotional intensity and text quality with human raters, finding effects that vary with the emotion and steering strength \cite{diallo2026style}. These findings motivate assessing emotional movement and readable expression separately. They also clarify what remains to be established for AFFECTOR: neither its inward/outward axis nor the correspondence between its RMS-scaled controls and a reader's emotional judgment has been validated by those studies.
 
-The practical need for bounded generation follows from evidence that emotional intervention can change more than tone. E-STEER studies effects on reasoning, generation, safety, and agent behavior and reports non-monotonic relationships \cite{sun2026emotion}. Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on architecture and extraction method \cite{jeong2026emotion}. AFFECTOR therefore treats repetition, truncation, and failed generations as outcomes to record, rather than assuming that stronger intervention produces a closer emotional match. Its rewrite styles make the competing requirements of emotional transformation and content preservation explicit.
+This uncertainty concerns the intervention itself as well as its evaluation. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition, while Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models \cite{shu2026emotion,wang2026readbest}. Thus, recognizing an emotion in a representation does not by itself determine where to intervene to change it. The partially coupled supportive-empathy directions reported by Ha and colleagues further raise the question of whether changing one emotional dimension also moves another \cite{ha2026empathy}. Research on automated vector extraction and context-dependent strength offers ways to investigate these problems, with benefits that depend on the task and intended trade-off \cite{cui2025pas,ferrando2025dsas}. AFFECTOR exposes layer choices and coordinate mixtures as experimental controls; testing their interactions and potential adaptation remains part of the research agenda.
 
-Choosing an evaluator and choosing an intervention site are also separate research problems. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition \cite{shu2026emotion}. Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models \cite{wang2026readbest}. Meanwhile, Ha and colleagues find partially coupled supportive-empathy directions \cite{ha2026empathy}, Cui and Chen automate vector extraction with task-dependent benefits \cite{cui2025pas}, and Ferrando and colleagues study context-dependent strength to balance toxicity mitigation and utility \cite{ferrando2025dsas}. Together, this work motivates testing layer choice, cross-axis interference, and adaptation. The current coordinate mixture is a simpler experimental control, rather than an implementation of those learned selectors or gates.
+When these interventions are applied throughout a document, their effects must also be considered over the course of generation. E-STEER reports non-monotonic relationships between emotional interventions and outcomes in reasoning, generation, safety, and agent behavior \cite{sun2026emotion}. Similarly, Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on the architecture and extraction method \cite{jeong2026emotion}. These results explain why greater intervention strength cannot simply be equated with a closer emotional match. In response to this practical difficulty, AFFECTOR combines bounded generation with records of repetition, truncation, and failed attempts. Its rewrite styles also let the user state how much content preservation should constrain the transformation, making that trade-off explicit when reviewing an output.
 
-The report consequently focuses on the implemented transformation pipeline and its observable limits. The application does not automatically score each generated passage and iteratively minimize its emotional distance; verification is a separate user-invoked operation. SteerScope's joint evaluation of efficacy, side effects, generalization, and data dependence supplies an appropriate broader evaluation agenda \cite{yang2026steerscope}. Its finding that activation methods do not dominate prompt steering in the tested trade-offs reinforces the need for prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Those comparisons remain research requirements, not completed evidence of AFFECTOR's effectiveness.
+Taken together, these considerations define the scope of the present report: the implemented transformation pipeline, the choices it exposes, and the limits visible in its operation. Emotional verification is a separate user-invoked step; the application does not automatically score each passage and iteratively minimize its distance from the target. A broader assessment must therefore compare the requested progression with independently judged outcomes. SteerScope provides a useful framework by evaluating efficacy alongside side effects, generalization, and dependence on steering data \cite{yang2026steerscope}. Its finding that activation methods do not dominate prompt steering in the tested trade-offs supports including prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Such comparisons would establish whether AFFECTOR's controls offer a reliable advantage for particular writing tasks; they remain research requirements rather than completed evidence of effectiveness.
 
 \section{Representation and
 alignment}\label{representation-and-alignment}
@@ -446,7 +446,7 @@ and limitations are stated separately in this report.
 \section*{Suggested citation}
 William Taggart. 2026. AFFECTOR: Coordinate-Guided Text Rewriting
 through Transformer Activation Interventions. Technical report. Modern
-and Classical Languages and Literatures, Virginia Tech. November 13, 2026.
+and Classical Languages and Literatures, Virginia Tech. November 14, 2026.
 \url{https://huggingface.co/will-taggart/AFFECTOR/blob/main/docs/TECHNICAL_PAPER.md}
 
 \end{document}

paper/PAPER-SHA256SUMS.txt

@@ -1,10 +1,10 @@
 322c798131b9aadcf7ada359c42f32c22adc44ab0dbd7e066c73cbe77ecfdf55  SCREENSHOTS.json
 d14edda139a36c0bac8b56d02e5cda63ab7558c6db1424f1087189a9429bfaf0  LITERATURE_SOURCES.md
-f4ffb7647cd04fa191e9b516485e08ea04fcf026d44a309cd667c7c817572458  AFFECTOR.tex
+95e51e9735f4fa3f8a1443ec37289e4beb0c8a23b09ec12829ded52f2c9dd053  AFFECTOR.tex
 9257f38423bd7e496f1a0916c0fbaefbfb32ab544a684ca9a411508201f0535e  references.bib
-27090c2045d62c461ab3c34410bda3d0fa7c4fc7d5656af720cd6fba5d3e99d2  source.md
-27090c2045d62c461ab3c34410bda3d0fa7c4fc7d5656af720cd6fba5d3e99d2  TECHNICAL_PAPER.md
-c8f7bbc77f3fc600efd5075b9a9230e29cae4f46d0b850721a0f26dc9651c977  AFFECTOR.pdf
+97a2a8a88f99be148a9b8a5774cd276cc38317712132ba024bd0b73b88a05760  source.md
+97a2a8a88f99be148a9b8a5774cd276cc38317712132ba024bd0b73b88a05760  TECHNICAL_PAPER.md
+632798acebaa3cbc687514ac8bdc029dd5dfd6a871bbb3ae7b9f323f5f6d616e  AFFECTOR.pdf
 d78577482f74cba4163db18d88af82786c923d72cd09607bde62cb747c645c5d  AFFECTOR.bbl
 eed8b979a5a299b42e925d9027332e7df876400dcef046f2cfe26beef91d1189  render_figures.py
 7f1c5ae6a0a8d61880debe26cc0b0a91ac74dd51841495920ae0cdd7ed23f83f  README.md
@@ -24,4 +24,4 @@ a6a497a38f4ba9de8088f410535bee13dd862d1a97f4175d205df768e1d2c1c9  figures/illust
 8d267fd6d50e2146279f1f33e07f88a3cbf76d023dd9ee36b29e390ac0473cd1  figures/screenshot-02.png
 759dabd81a2ce67f01a3ec96a01b014468402baee5774792fccd5b7fe53af80a  figures/screenshot-03.png
 ff1ace53310f4f8ea57dedb7a202edd7404886091ed34b1bcc70cbb6c93b2779  figures/screenshot-01.png
-f224f098413cc47d60d52dd48283b595f815574cb709c2357385d930a38bc701  AFFECTOR-source.zip
+7a96304eae89114abc39be017b5163b88388efbfbb0ef017fae59ecfaa0721c6  AFFECTOR-source.zip

paper/TECHNICAL_PAPER.md

@@ -5,7 +5,7 @@
 **William Taggart**  
 Modern and Classical Languages and Literatures, Virginia Tech  
 Email: wtaggart@vt.edu  
-November 13, 2026 · Implementation report · Desktop version 0.1.0
+November 14, 2026 · Implementation report · Desktop version 0.1.0
 
 ## Abstract
 
@@ -13,21 +13,21 @@ AFFECTOR investigates whether a prescribed emotional trajectory can influence th
 
 ## Purpose and Research Context
 
-A document conveys more than propositions: lexical choice, rhythm, imagery, focus, and narrative stance contribute to its emotional character. AFFECTOR investigates whether a user-defined emotional progression can guide a rewrite while leaving room to choose how much of the original content to preserve. It represents that progression as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. Such a curve specifies a coarse target rather than a unique text, so emotional fit must be considered alongside readability, content change, and output length.
+A document conveys more than propositions. Its lexical choices, rhythm, imagery, focus, and narrative stance also shape how a reader feels as the text unfolds. AFFECTOR begins with a practical question about that progression: can a writer specify a desired emotional path and use it to guide a rewrite, while choosing how much of the original content to preserve? To make this question tractable, the application represents the path as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. This representation supplies a coarse target for experimentation, leaving the model to realize it in language. The resulting text must therefore be judged both by its emotional fit and by its readability, content changes, and length.
 
-Activation steering provides one mechanism for pursuing this target. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). AFFECTOR uses contrastive calibration to map emotional coordinates into changes to a frozen model's internal states during local inference. Its API backend instead describes the coordinates in instructions to a hosted model. These two routes support comparison between numerical intervention and prompt-based control, but their strengths and outcomes cannot be treated as equivalent.
+Activation steering offers a way to connect this target to the process of generation. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). Building on this contrastive approach, AFFECTOR maps emotional coordinates into changes to a frozen model's internal states during local inference. The application also provides an API route, in which the coordinates are expressed as instructions to a hosted model. Together, these routes make it possible to investigate numerical intervention alongside prompt-based control, although their settings and outcomes are not directly interchangeable.
 
 ![Document workspace with an example input, rewrite controls, and an empty output pane before generation.](figures/screenshot-01.png)
 
 *Document workspace with an example input, rewrite controls, and an empty output pane before generation.*
 
-The choice to expose emotional coordinates is informed by work connecting model representations with human judgments. Wu and colleagues relate model emotion spaces to human valence/arousal ratings and report concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues assess emotional intensity and text quality with human raters, finding effects that depend on the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These studies motivate separate assessments of emotional movement and readable expression. They do not validate AFFECTOR's inward/outward axis or establish a correspondence between their intervention scales and its RMS-scaled controls.
+For either route, the usefulness of the coordinates depends on their relationship to human interpretations of the output. Work by Wu and colleagues connects model emotion spaces with human valence/arousal ratings and reports concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues extend this evaluative perspective by examining emotional intensity and text quality with human raters, finding effects that vary with the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These findings motivate assessing emotional movement and readable expression separately. They also clarify what remains to be established for AFFECTOR: neither its inward/outward axis nor the correspondence between its RMS-scaled controls and a reader's emotional judgment has been validated by those studies.
 
-The practical need for bounded generation follows from evidence that emotional intervention can change more than tone. E-STEER studies effects on reasoning, generation, safety, and agent behavior and reports non-monotonic relationships ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). AFFECTOR therefore treats repetition, truncation, and failed generations as outcomes to record, rather than assuming that stronger intervention produces a closer emotional match. Its rewrite styles make the competing requirements of emotional transformation and content preservation explicit.
+This uncertainty concerns the intervention itself as well as its evaluation. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition, while Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3); [Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Thus, recognizing an emotion in a representation does not by itself determine where to intervene to change it. The partially coupled supportive-empathy directions reported by Ha and colleagues further raise the question of whether changing one emotional dimension also moves another ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)). Research on automated vector extraction and context-dependent strength offers ways to investigate these problems, with benefits that depend on the task and intended trade-off ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3); [Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). AFFECTOR exposes layer choices and coordinate mixtures as experimental controls; testing their interactions and potential adaptation remains part of the research agenda.
 
-Choosing an evaluator and choosing an intervention site are also separate research problems. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3)). Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Meanwhile, Ha and colleagues find partially coupled supportive-empathy directions ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)), Cui and Chen automate vector extraction with task-dependent benefits ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3)), and Ferrando and colleagues study context-dependent strength to balance toxicity mitigation and utility ([Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). Together, this work motivates testing layer choice, cross-axis interference, and adaptation. The current coordinate mixture is a simpler experimental control, rather than an implementation of those learned selectors or gates.
+When these interventions are applied throughout a document, their effects must also be considered over the course of generation. E-STEER reports non-monotonic relationships between emotional interventions and outcomes in reasoning, generation, safety, and agent behavior ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Similarly, Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on the architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). These results explain why greater intervention strength cannot simply be equated with a closer emotional match. In response to this practical difficulty, AFFECTOR combines bounded generation with records of repetition, truncation, and failed attempts. Its rewrite styles also let the user state how much content preservation should constrain the transformation, making that trade-off explicit when reviewing an output.
 
-The report consequently focuses on the implemented transformation pipeline and its observable limits. The application does not automatically score each generated passage and iteratively minimize its emotional distance; verification is a separate user-invoked operation. SteerScope's joint evaluation of efficacy, side effects, generalization, and data dependence supplies an appropriate broader evaluation agenda ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs reinforces the need for prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Those comparisons remain research requirements, not completed evidence of AFFECTOR's effectiveness.
+Taken together, these considerations define the scope of the present report: the implemented transformation pipeline, the choices it exposes, and the limits visible in its operation. Emotional verification is a separate user-invoked step; the application does not automatically score each passage and iteratively minimize its distance from the target. A broader assessment must therefore compare the requested progression with independently judged outcomes. SteerScope provides a useful framework by evaluating efficacy alongside side effects, generalization, and dependence on steering data ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs supports including prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Such comparisons would establish whether AFFECTOR's controls offer a reliable advantage for particular writing tasks; they remain research requirements rather than completed evidence of effectiveness.
 
 ## Representation and alignment
 

paper/source.md

@@ -5,7 +5,7 @@
 **William Taggart**  
 Modern and Classical Languages and Literatures, Virginia Tech  
 Email: wtaggart@vt.edu  
-November 13, 2026 · Implementation report · Desktop version 0.1.0
+November 14, 2026 · Implementation report · Desktop version 0.1.0
 
 ## Abstract
 
@@ -13,21 +13,21 @@ AFFECTOR investigates whether a prescribed emotional trajectory can influence th
 
 ## Purpose and Research Context
 
-A document conveys more than propositions: lexical choice, rhythm, imagery, focus, and narrative stance contribute to its emotional character. AFFECTOR investigates whether a user-defined emotional progression can guide a rewrite while leaving room to choose how much of the original content to preserve. It represents that progression as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. Such a curve specifies a coarse target rather than a unique text, so emotional fit must be considered alongside readability, content change, and output length.
+A document conveys more than propositions. Its lexical choices, rhythm, imagery, focus, and narrative stance also shape how a reader feels as the text unfolds. AFFECTOR begins with a practical question about that progression: can a writer specify a desired emotional path and use it to guide a rewrite, while choosing how much of the original content to preserve? To make this question tractable, the application represents the path as an ordered curve of valence, inward/outward emotional direction, and intensity across the document. This representation supplies a coarse target for experimentation, leaving the model to realize it in language. The resulting text must therefore be judged both by its emotional fit and by its readability, content changes, and length.
 
-Activation steering provides one mechanism for pursuing this target. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). AFFECTOR uses contrastive calibration to map emotional coordinates into changes to a frozen model's internal states during local inference. Its API backend instead describes the coordinates in instructions to a hosted model. These two routes support comparison between numerical intervention and prompt-based control, but their strengths and outcomes cannot be treated as equivalent.
+Activation steering offers a way to connect this target to the process of generation. Turner and colleagues construct intervention vectors from contrasting prompts, while Panickssery and colleagues average activation differences across positive and negative behavioral examples ([Turner et al., 2023](https://arxiv.org/abs/2308.10248v1); [Panickssery et al., 2024](https://arxiv.org/abs/2312.06681v4)). Building on this contrastive approach, AFFECTOR maps emotional coordinates into changes to a frozen model's internal states during local inference. The application also provides an API route, in which the coordinates are expressed as instructions to a hosted model. Together, these routes make it possible to investigate numerical intervention alongside prompt-based control, although their settings and outcomes are not directly interchangeable.
 
 ![Document workspace with an example input, rewrite controls, and an empty output pane before generation.](figures/screenshot-01.png)
 
 *Document workspace with an example input, rewrite controls, and an empty output pane before generation.*
 
-The choice to expose emotional coordinates is informed by work connecting model representations with human judgments. Wu and colleagues relate model emotion spaces to human valence/arousal ratings and report concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues assess emotional intensity and text quality with human raters, finding effects that depend on the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These studies motivate separate assessments of emotional movement and readable expression. They do not validate AFFECTOR's inward/outward axis or establish a correspondence between their intervention scales and its RMS-scaled controls.
+For either route, the usefulness of the coordinates depends on their relationship to human interpretations of the output. Work by Wu and colleagues connects model emotion spaces with human valence/arousal ratings and reports concept-based steering across linguistic-cultural settings ([Wu et al., 2025](https://arxiv.org/abs/2506.13978v1)). Diallo and colleagues extend this evaluative perspective by examining emotional intensity and text quality with human raters, finding effects that vary with the emotion and steering strength ([Diallo et al., 2026](https://arxiv.org/abs/2601.21505v1)). These findings motivate assessing emotional movement and readable expression separately. They also clarify what remains to be established for AFFECTOR: neither its inward/outward axis nor the correspondence between its RMS-scaled controls and a reader's emotional judgment has been validated by those studies.
 
-The practical need for bounded generation follows from evidence that emotional intervention can change more than tone. E-STEER studies effects on reasoning, generation, safety, and agent behavior and reports non-monotonic relationships ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). AFFECTOR therefore treats repetition, truncation, and failed generations as outcomes to record, rather than assuming that stronger intervention produces a closer emotional match. Its rewrite styles make the competing requirements of emotional transformation and content preservation explicit.
+This uncertainty concerns the intervention itself as well as its evaluation. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition, while Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3); [Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Thus, recognizing an emotion in a representation does not by itself determine where to intervene to change it. The partially coupled supportive-empathy directions reported by Ha and colleagues further raise the question of whether changing one emotional dimension also moves another ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)). Research on automated vector extraction and context-dependent strength offers ways to investigate these problems, with benefits that depend on the task and intended trade-off ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3); [Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). AFFECTOR exposes layer choices and coordinate mixtures as experimental controls; testing their interactions and potential adaptation remains part of the research agenda.
 
-Choosing an evaluator and choosing an intervention site are also separate research problems. Shu and colleagues find uneven causal organization across emotion categories in a study of sparse features and emotion recognition ([Shu et al., 2026](https://arxiv.org/abs/2604.25866v3)). Wang and colleagues show that strong emotion-probing accuracy need not identify the best intervention layers in omni-modal models ([Wang et al., 2026](https://arxiv.org/abs/2609.22135v1)). Meanwhile, Ha and colleagues find partially coupled supportive-empathy directions ([Ha et al., 2026](https://arxiv.org/abs/2609.15654v1)), Cui and Chen automate vector extraction with task-dependent benefits ([Cui et al., 2025](https://arxiv.org/abs/2509.22739v3)), and Ferrando and colleagues study context-dependent strength to balance toxicity mitigation and utility ([Ferrando et al., 2025](https://arxiv.org/abs/2512.03661v2)). Together, this work motivates testing layer choice, cross-axis interference, and adaptation. The current coordinate mixture is a simpler experimental control, rather than an implementation of those learned selectors or gates.
+When these interventions are applied throughout a document, their effects must also be considered over the course of generation. E-STEER reports non-monotonic relationships between emotional interventions and outcomes in reasoning, generation, safety, and agent behavior ([Sun et al., 2026](https://arxiv.org/abs/2604.00005v1)). Similarly, Jeong's small-model comparison finds both coherent emotional changes and degeneration, including repetition, depending on the architecture and extraction method ([Jeong, 2026](https://arxiv.org/abs/2604.04064v1)). These results explain why greater intervention strength cannot simply be equated with a closer emotional match. In response to this practical difficulty, AFFECTOR combines bounded generation with records of repetition, truncation, and failed attempts. Its rewrite styles also let the user state how much content preservation should constrain the transformation, making that trade-off explicit when reviewing an output.
 
-The report consequently focuses on the implemented transformation pipeline and its observable limits. The application does not automatically score each generated passage and iteratively minimize its emotional distance; verification is a separate user-invoked operation. SteerScope's joint evaluation of efficacy, side effects, generalization, and data dependence supplies an appropriate broader evaluation agenda ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs reinforces the need for prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Those comparisons remain research requirements, not completed evidence of AFFECTOR's effectiveness.
+Taken together, these considerations define the scope of the present report: the implemented transformation pipeline, the choices it exposes, and the limits visible in its operation. Emotional verification is a separate user-invoked step; the application does not automatically score each passage and iteratively minimize its distance from the target. A broader assessment must therefore compare the requested progression with independently judged outcomes. SteerScope provides a useful framework by evaluating efficacy alongside side effects, generalization, and dependence on steering data ([Yang et al., 2026](https://arxiv.org/abs/2610.07722v1)). Its finding that activation methods do not dominate prompt steering in the tested trade-offs supports including prompt-only and unsteered baselines, blinded emotional and quality ratings, and complete failure accounting. Such comparisons would establish whether AFFECTOR's controls offer a reliable advantage for particular writing tasks; they remain research requirements rather than completed evidence of effectiveness.
 
 ## Representation and alignment