Advanced search

Find records

with all of the words

with the exact phrase

with at least one of the words

without the words

where my words occur anywhere in the record in the title of the record

Narrow down

Return records authored by

Return records published in

Return records dated between

Return records of type

Search tips Browse subjects

6 results

Preprint

A Best Match Is Not a Twin: What Feature Similarity Between State Space Models and Transformers Certifies, Why the Null and the Shared Inputs Decide the Verdict, and the Controls That Would Show Two Architectures Learned the Same Thing

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Selective state space models such as Mamba compress the past into a fixed-size state, while Transformers attend over the whole context, and a growing set of studies asks whether the two learn the same internal features. A 2026…

Cite Related research LOR-00142 · 12 views · 11 downloads Select “A Best Match Is Not a Twin: What Feature Similarity…” for export

PDF403 KB

Preprint

Newly Important Is Not Newly Built: What Reasoning Post-Training Adds Inside a Language Model, Why the Evidence Separates Representation From Wiring From Deployment, and the Cross-Model Tests That Would Show an Attention Head Is New

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Life of Research Reasoning models are built by post-training a pretrained model with supervised fine-tuning on teacher traces, with reinforcement learning against a verifier, or with both, and a running dispute asks whether this builds new…

Cite Related research LOR-00125 · 17 views · 15 downloads Select “Newly Important Is Not Newly Built: What Reasoning…” for export

PDF378.2 KB

Journal article

Repair Without a Repairer: Five Mechanisms Called Self-Repair in Neural Networks, Why Much of the Measured Restoration Is Restored Scale, and What Would License the Claim That a Network Restores Its Own Performance

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. "Self-repair" is used in at least three research communities that do not cite one another. In transformer interpretability it names the Hydra effect: when one component is ablated, later components shift their output and partly restore the…

Cite Related research LOR-00104 · 16 views · 52 downloads Select “Repair Without a Repairer: Five Mechanisms Called…” for export

PDF411.6 KB

Journal article

Circuits Are Estimates, Not Mechanisms: A Recovered Subgraph Is One Draw From a Class Set by the Ablation, the Dataset and the Grain, the Results That Defend Circuits Hold at the Level of That Class, and What Ablation Evidence Licenses at Each Level

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. A circuit in a language model is usually presented as the mechanism by which the model performs a task: a sparse subgraph, found by ablation or patching, that reproduces the behaviour when everything else is knocked out. A second body of work…

Cite Related research LOR-00098 · 17 views · 30 downloads Select “Circuits Are Estimates, Not Mechanisms: A Recovered…” for export

PDF395.2 KB

Journal article

A Control Knob Is Not a Concept: What a Working Steering Vector Licenses, Why Its Equivalence Class Is Set by the Readout That Scores It, and the Controls That Would Upgrade Control to Representation

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Adding a vector to a language model's residual stream and watching a behaviour change has become the field's most common argument that the model represents a concept along that vector. Refusal directions, persona vectors, truth directions and…

Cite Related research LOR-00099 · 17 views · 27 downloads Select “A Control Knob Is Not a Concept: What a Working Steering…” for export

PDF393 KB

Journal article

Dictionaries Without Ontologies: What a Sparse Autoencoder Latent Is Evidence For, Why Decodability, Use and Reproducibility Do Not Add Up to Canonicity, and the Identifiability Premise No Real-Model Result Has Checked

Pranay Mahendrakar 2026

(c) 2026 Pranay Mahendrakar. Licensed under CC BY 4.0. Sparse autoencoders are now the default instrument for decomposing language model activations, and their latents are routinely reported as the model's features: units the network itself uses, which a wide enough dictionary has pulled out of…

Cite Related research LOR-00096 · 34 views · 23 downloads Select “Dictionaries Without Ontologies: What a Sparse Autoencoder…” for export

PDF405.3 KB

Search tips

Type this	To find
`neural networks`	Records containing both words, anywhere (titles count most)
`"language models"`	The exact phrase
`memory -survey`	Leave out records that mention a word
`hallucination OR confabulation`	Either word
`author:"Ada Lovelace"`	Records by an author
`title:benchmark`	A word in the title
`allintitle: agent memory`	All of the words in the title
`source:"Nature"`	Records published in a journal
`keyword:"machine learning"`	Records tagged with a subject
`year:2022-2026`	Published in a range of years (or year:2026)
`since:2025`	Published in or after a year (until:2020 for before)
`type:dataset`	Only one type: article, preprint, dataset, thesis, report, software…
`doi:10.5281/zenodo.123456`	The record with a DOI
`related:LOR-00012`	Research related to a record

Operators combine: "open access" author:Mahendrakar since:2025 -preprint. Words match their variants (model finds models) and accents are ignored. Press / anywhere to jump to the search box. Software and agents can use the same syntax through the JSON search API.

Cite

APA

BibTeX

Import into a reference manager: BibTeX · RIS (Zotero, Mendeley, EndNote)